ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)

RNS INSTITUTE OF TECHNOLOGY

Dr. VISHNUVARDHAN ROAD, CHANNASANDRA, BENGALURU -560 098

Department of Information Science and Engineering

VISION of the College

Building RNSIT into a World - Class Institution

MISSION of the College

To impart high quality education in Engineering, Technology and Management with a difference, enabling
students to excel in their career by

1. Attracting quality Students and preparing them with a strong foundation in fundamentals so as to
achievedistinctions in various walks of life leading to outstanding contributions.
2. Imparting value based, need based, and choice based and skill based professional education to the
aspiringyouth and carving them into disciplined, World class Professionals with social responsibility.
3. Promoting excellence in Teaching, Research and Consultancy that galvanizes academic consciousness
among Faculty and Students.
4. Exposing Students to emerging frontiers of knowledge in various domains and make them suitable
forIndustry, Entrepreneurship, Higher studies, and Research & Development.
5. Providing freedom of action and choice for all the Stake holders with better visibility.

VISION of the Department

Fostering winning professionals of Strong Informative Potentials.

MISSION of the Department

Imparting high quality education in the area of Information Science so as to graduate the students
with good fundamentals, "Information System Integration", "Software Creation" capability & suitably train
them to thrive in Industries, higher schools of learning and R & D centers with a comprehensive
perspective.

PROGRAM EDUCATIONAL OBJECTIVES (PEOs)

ISE Graduates within three-four years of graduation will have
• PEO1: Acquired the fundamentals of computers and applied knowledge of Information Science &
Engineering and continue to develop their technical competencies by problem solving using
programming.
• PEO2: Ability to formulate problems, attained the Proficiency to develop system/application

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 1

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
software in a scalable and robust manner with various platforms, tools and frameworks to provide
cost effective solutions.
• PEO3: Obtained the capacity to investigate the necessities of the software Product, adapt to
technological advancement, promote collaboration and interdisciplinary activities, Protecting
Environment and developing Comprehensive leadership.
• PEO4: Enabled to be employed and provide innovative solutions to real-world problems across
different domains.
• PEO5: Possessed communication skills, ability to work in teams, professional ethics, social
responsibility, entrepreneur and management, to achieve higher career goals, and pursue higher
studies.

PROGRAM OUTCOMES (POs)

Engineering Graduates will be able to:
• PO1: Engineering knowledge: Apply the knowledge of mathematics, science, engineering
fundamentals, and an engineering specialization for the solution of complex engineering problems
• PO2: Problem analysis: Identify, formulate, research literature, and analyze complex
engineering problems reaching substantiated conclusions using first principles of mathematics,
natural sciences and engineering sciences.
• PO3: Design/development of solutions: Design solutions for complex engineering problems and
design system components or processes that meet the specified needs with appropriate
consideration for public health and safety, and cultural, societal, and environmental
considerations.
• PO4: Conduct investigations of complex problems: Use research-based knowledge and research
methods including design of experiments, analysis and interpretation of data, and synthesis of
the information to provide valid conclusions.
• PO5: Modern tool usage: Create, select, and apply appropriate techniques, resources, and
modern engineering and IT tools, including prediction and modeling to complex engineering
activities, with an understanding of the limitations.
• PO6: The engineer and society: Apply reasoning informed by the contextual knowledge to
assess Societal, health, safety, legal and cultural issues and the consequent responsibilities relevant
to the professional engineering practice.
• PO7: Environment and sustainability: Understand the impact of the professional engineering
solutions in societal and environmental contexts, and demonstrate the knowledge of, and need for
sustainable development.
• PO8: Ethics: Apply ethical principles and commit to professional ethics and responsibilities and norms
of the engineering practice.
• PO9: Individual and team work: Function effectively as an individual, and as a member or leader
in diverse teams, and in multidisciplinary settings.
• PO10: Communication: Communicate effectively on complex engineering activities with the
engineering community and with the society at large, such as, being able to comprehend and write
effective reports and design documentation, make effective presentations, and give and receive

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 2

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
clear instructions.
• PO11: Project management and finance: Demonstrate knowledge and understanding of the
engineering and management principles and apply these to one’s own work, as a member and
leader in a team, to manage projects and in multidisciplinary environments.
• PO12: Life-long learning: Recognize the need for, and have the preparation and ability to engage
in independent and life-long learning in the broadest context of technological change.

PROGRAM SPECIFIC OUTCOMES (PSOs)

ISE Graduates will have

• PSO1 – Problem Solving Abilities: Ability to demonstrate the fundamental and theoretical concepts,
analyze the real time problems and develop customized software solutions by applying the
knowledge of mathematics and algorithmic techniques.

• PSO2 – Applied Engineering Skills: Enable creative thinking, Ability to apply standard practices and
strategies, technical skills in software design, development, integration of systems and management
forimproving the security, reliability and survivability of the infrastructure.
• PSO3 – General Expertise and Higher Learning: Ability to exchange knowledge effectively
demonstrate the ability of team work, documentation skills, professional ethics, entrepreneurial skills
and continuing higher education in the field of Information technology.

COURSE OBJECTIVES OF AI&ML

This course will enable students to
• Explain Artificial Intelligence and Machine Learning
• Illustrate AI and ML algorithm and their use in appropriate applications

COURSE OUTCOMES OF AI&ML

After studying this course, students will be able to:
• CO1: Analyze and formalize the real world problems as a state space, graph, design heuristics
and select amongst different search or game based techniques to solve them.
• CO2: Attain the capability to represent and solve various real life problems using logic based
techniques and use this to perform inference or reasoning.
• CO3: Investigate Concept Learning and Apply Decision Tree concepts on various data sets.
• CO4: Solve the problems related to Artificial Neural Networks.
• CO5: Analyze Bayesian learning algorithm to classify different data sets.
• CO6: Illustrate the concepts of Instance based learning and Reinforcement learning
algorithms.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 3

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING
Module I
Topics: What is artificial intelligence?, Problems, problem spaces and search, Heuristic search
techniques

Chapter: What Is Artificial Intelligence?

A broad field that means different things to different people

✓ Artificial intelligence (AI) is the study of how to make computers do things which, at the
moment, people do better – Rich and Knight, 1991
✓ The study of the computations that make it possible to perceive, reason and act – Patrick Henry
Winston, 1992
✓ AI is a branch of computer science deals with automation of intelligent behavior – Luger &
Stubblefield, 1993
✓ A field of study that seeks to explain and emulate the intelligent behavior in terms of
computational processes – Schalkoff, 1990
✓ The automation of activities that we associate with human thinking – Bellman, 1978
Activities Means: Decision making, problem solving, learning ….
✓ The study of techniques to make the computers to exhibit some kind of intelligence
✓ It is the science and engineering of making intelligent machines, especially intelligent computer
programs.
It is related to the similar task of using computers to understand human intelligence
i.e., AI deals with the techniques to make computers to exhibit some kind of intelligence.

1.1 THE AI PROBLEMS

✓ Much of the early work in the field focused on formal tasks, such as game playing and theorem
proving.
✓ Samuel wrote a checkers-playing program that not only played games with opponents but also
used its experience at those games to improve its later performance. Chess also received a good
deal of attention.
✓ The Logic Theorist was an early attempt to prove mathematical theorems. It was able to prove
several theorems. Gelernter’s theorem explored another area of mathematics: geometry. Game
playing and theorem proving share the property that people who do them well are considered to
be displaying intelligence.
✓ In spite of this, it appeared initially that computers could perform well at those tasks simply by
being fast at exploring a large number of solution paths and then selecting the best one. It was
thought that this process required very little knowledge and could therefore be programmed
easily.
✓ Another early work in AI focused is on the sort of problems solving called commonsense
reasoning. It includes reasoning about physical objects and their relationships to each other. To

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 4

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
investigate this sort of reasoning, Newell, Shaw, and Simon built the General Problem Solver
(GPS) - here only simple tasks were selected.
✓ AI research progressed and techniques for handling larger amounts of world knowledge were
developed. These include perception – vision and speech, natural language understanding,
and problems solving in specialized domains such as medical diagnosis and chemical analysis
and are referred as Mundane tasks.
✓ In addition to mundane tasks, people can also perform specialized tasks in which carefully
acquired expertise is necessary. Examples of such tasks include engineering design, scientific
discovery, medical diagnosis and financial planning. Figure below lists some of the tasks that
are targets of work in AI.
✓ As a result, the problem areas where AI is now flourishing most as a practical discipline are
primarily the domains that require only specialized expertise without the assistance of
commonsense knowledge.

Some of the Task Domains of AI

✓ Mundane Tasks
– Perception
• Vision
• Speech
– Natural Languages
• Understanding
• Generation
• Translation
– Common sense reasoning
– Robot Control
✓ Formal Tasks
– Games Playing
• Chess
• Backgammon
• Checkers - Go
– Mathematics
• Geometry
• Logic
• Integral calculus
• Theorem Proving
• General Problem Solving
✓ Expert Tasks (require specialized skills and training)
SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 5
 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
– Engineering
• Design
• Fault finding
• Manufacturing planning
– Scientific Analysis
– Medical Diagnosis
– Financial Analysis
Note: AI is concerned with automating both Mundane and Expert tasks

1.2 THE UNDERLYING ASSUMPTION

The heart of research in artificial intelligence lies what Newell and Simon [1976] call the physical
symbol system hypothesis. They define a Physical symbol system as follows:

✓ A physical symbol system consist of a set of entities called symbols, which are physical
patterns that can occur as components of another type of entity called an expression or symbol
structure. Thus, a symbol structure is composed of a number of instances or tokens of symbols
related in some physical way. At any instant of time the system will contain a collection of these
symbol structures.
✓ In addition to these structures, the system also contains a collection of processes that operate on
expressions to produce other expressions: processes of creation, modification, reproduction
and destruction. A physical symbol system is a machine that produces through time, an
evolving collection of symbol structures.
✓ The physical Symbol System Hypothesis can be stated as – A physical symbol system has the
necessary and sufficient means for general intelligent action.
✓ The truth of this hypothesis can be determined only by experimentation. Computers provide
the perfect medium for this experimentation since they can be programmed to simulate any
physical symbol system. This ability of computers to serve as arbitrary symbol manipulators was
noticed very early in the history of computing by Lady Lovelace about Babbage’s proposed
Analytical Engine in 1842.
✓ The operating mechanism can even be thrown into action independently of any object to operate
upon. As it has become increasingly easy to build computing machines, so it has become
increasingly possible to conduct empirical investigations of the physical symbol system
hypothesis. In each such investigation, a particular task that might be regarded as requiring
intelligence is selected. A program to perform the task is proposed and then tested. We have not
been completely successful at creating programs that perform all the selected tasks.
✓ Evidence in support of the physical symbol system hypothesis has come not only from areas such
as game playing, but also from areas such as visual perception, where it is more tempting to
suspect the influence of sub-symbolic processes. However, sub-symbolic models such as neural
networks are beginning to challenge symbolic ones at such low-level tasks.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 6

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
1.3 WHAT IS AN AI TECHNIQUE?

Artificial intelligence problems span a very broad spectrum. There are techniques that are appropriate for
the solution of a variety of these problems.

One of the few hard and fast results to come out of the first three decades of AI research is that
intelligence requires knowledge. The knowledge possesses some properties as follows:

✓ It is voluminous
✓ It is hard to characterize accurately
✓ It is constantly changing
✓ It differs from data being organized in a way that corresponds to the ways it will be used

Here, it is concluded that an AI technique is a method that exploits knowledge that should be
represented in such a way that:

✓ The knowledge captures generalizations means that, it is not necessary to represent separately
each individual situation. Instead, situations that share important properties are grouped together.
If knowledge does not have this property, excessive amounts of memory and updating will be
require, then called as data rather than knowledge.
✓ It can be understood by people who must provide it. Although for many programs, the bulk of the
data can be acquired automatically, for example by taking readings from a variety of instruments.
✓ It can easily be modified to correct errors and to reflect changes in the world and in our world
view.
✓ It can be used in many situations even if it is not totally accurate or complete.
✓ It can be used to help overcome its own complete volume by helping to narrow the range of
possibilities.

Although AI techniques must be designed in keeping with these constraints imposed by Al problems, there
is some degree of independence between problems and problem-solving techniques. It is possible to solve Al
problems without using AI techniques and it is possible to apply AI techniques to the solution of non-Al
problems.

In order to characterize AI techniques as problem-independent way as possible, let us consider some

problems.

1.3.1 Tic-Tac-Toe problem

Here, presented a series of three programs to play tic-tac-toe. The programs in this series increase in:

✓ The clarity of their knowledge

✓ Their complexity
✓ Their use of generalizations
✓ The extensibility of their approach. Thus, they move toward being representations of what we call
AI techniques.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 7

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Program 1

Data Structures:
Board A nine-element vector representing the board, where the elements of the vector correspond
to the board positions as follows:

1 2 3
4 5 6
7 8 9
An element contains the value 0 if the corresponding square is blank. 1 if it is filled with
an X, or 2 if it is filled with an O.

Movetable A large vector of 19,683 elements (39 ), each element of which is a nine-element vector.

The contents of this-vector are chosen specifically to allow the algorithm to work.

Algorithm: To make a move, do the following:

1. View the vector Board as a ternary (base three) number. Convert it to a decimal number.
2. Use the number computed in step 1 as an index into Movetable and access the vector stored
there.
3. The vector selected in step 2 represents the way the board will look after the move that should be
made. So set Board equal to that vector.

Comments:

This program is very efficient in terms of time. And, in theory, it could play an optimal game of tic-tac-
toe. But it has several disadvantages:

✓ It takes a lot of space to store the table of each move

✓ Have to do a lot of work specifying all the entries in the movetable.
✓ It is very unlikely that all the required movetable entries can be determined and entered
without any errors.
✓ If we want to extend the game, say to three dimensions, we would have to start from scratch,
since 327 board positions would have to be stored, occupies large memory.

Program 2

Data Structures:

Board A nine-element vector representing the board, as described for Program 1. But instead of
using the numbers 0, 1 or 2 in each element, we store 2 is blank, 3 for X and 5 for O.

Turn An integer indicating which move of the game is about to be played; 1 indicates the first
move, 9 the last.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 8

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Algorithm: The main algorithm uses three sub-procedures:

Make 2 Returns 5 if the center square of the board is blank, that is, if Board [5] = 2. Otherwise,
this function returns any blank Noncorner Square – 2, 4, 6, or 8.

Posswin (p) Returns 0 if player p cannot win on his next move; otherwise, it returns the number of the
square that constitutes a winning move. This function will enable the program both to win
and to block the opponent's win. Posswin operates by checking, one at a time, each of the
rows, columns, and diagonals. Because of the way values are numbered it can test an entire
row/column/diagonal to see if it is a possible win by multiplying the values of its squares
together. If the product is 18 (3 x 3 x 2), then X can win (means one of the squares is
empty). If the product is 50 (5 x 5 x 2), then O can win (means one of the squares is
empty). If we find a winning row, we determine which element is blank, and return the
number of that square.

Go (n) Makes a move in square n. This procedure sets Board[n] to 3 if Turn is odd, or 5 if Turn
is even. It also increments Turn by one.

The algorithm has a built-in strategy for each move. It makes the odd-numbered moves if it is playing X, the
even-numbered moves if it is playing O. The strategy for each turn is as follows:

Turn= 1 Go(1 ) (upper left corner).

Turn= 2 If Board [5] is blank, Go(5), else Go(1).
Turn= 3 If Board [9] is blank, Go(9), else Go(3).
Turn= 4 If Posswin(X) is not 0, then Go(Posswin(X)) [i.e., block opponent's win],
else Go(Make2).
Turn= 5 If Posswin(X) is not 0 then Go(Posswin(X))[i.e., win] else if Posswin(O) is not 0,
then Go(Posswin(O)) [i.e., block win], else if Board[7] is blank, then Go(7), else Go(3).
[Here the program is trying to make a fork.]
Turn= 6 If Posswin(O) is not 0 then Go (Posswin(O)), else if Posswin(X) is not 0, then
Go(Posswin(X)), else Go(Make2).
Turn= 7 If Posswin(X) is not 0 then Go(Posswin(X)). else if Posswin(O) is not 0,
then Go(Posswin(O)). Else go anywhere that is blank.
Turn= 8 If Posswin(O) is not 0 then Go(Posswin(O)), else if Posswin(X) is not 0, then
Go(Posswin(X)), else go anywhere that is blank.
Turn= 9 Same as Turn=7.
Program 2 – A

This program is identical to program 2 except for the one change in the representation of the board and is
as follows:
8 3 4

1 5 9
6 7 2

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 9

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Here, numbering of the board produces a magic square: all the rows, columns, and diagonals sum up to
15. This means that we can simplify the process of checking for a possible win. In addition to marking the
board as moves are made, we keep a list, for each player, of the squares in which he or she has played. To
check for a possible win for one player, we consider each pair of squares owned by that player and
compute the difference between 15 and the sum of the two squares. If this difference is not positive or
if it is greater than 9, then the original two squares were not collinear and so can be ignored. Otherwise,
if the square representing the difference is blank (1 to 9 squares), a move there will produce a win. This
shows how the choice of representation can have a major impact on the efficiency of a problem-solving
program.
Program 3

Data Structures

BoardPosition: A structure containing a nine-element vector representing the board, a list of board
positions that could result from the next move, and a number representing an estimate of how likely the
board position is to lead to an ultimate win for the player to move.
Algorithm:
To decide on the next move, look ahead at the board positions that result from each possible move. Decide
which position is best (as described below), make the move that leads to that position, and assign the rating of
that best move to the current position.

To decide which of a set of board positions is best, do the following for each of them:

1. See if it is a win. If so, call it the best by giving it the highest possible rating.
2. Otherwise, consider all the moves the opponent could make next. See which of them is worst for us
(by recursively calling this procedure). Assume the opponent will make that move. Whatever
rating that moves has, assign it to the node we are considering.
3. The best node is then the one with the highest rating.

This algorithm will look ahead at various sequences of moves in order to find a sequence that leads to a
win. It attempts to maximize the likelihood of winning, while assuming that the opponent will try to
minimize that likelihood. This algorithm is called the minimax procedure.

1.3.2 Question Answering problem

Here, programs that read in English text and then answer questions also stated in English, about that
text. This is more difficult to state formally and precisely what our problem is and what constitutes
correct solutions to it. For example, suppose that the input text were just the single sentence

Example 1:
Russia massed troops on the Czech border.
Then either of the following question-answering dialogues might occur with the POLITICS program:

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 10

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Dialogue 1
Q: Why did Russia do this?
A: Because Russia thought that it could take political control of Czechoslovakia by sending troops.
Q: What should the United States do?
A: The United States should intervene (interfere) militarily.

Dialogue 2
Q: Why did Russia do this?
A: Because Russia wanted to increase its political influence over Czechoslovakia.
Q: What should the United States do?
A: The United States should denounce (condemn) the Russian action in the United Nations.

✓ In the POLITICS program, answers were constructed by considering both the input text and a separate
model of the beliefs and actions of various political entities, including Russia. When the model is changed,
the system’s answers also change.
✓ The general point here is that defining what it means to produce a correct answer to a question may be very
hard.

Example 2:
Mary went shopping for a new coat. She found a red one she really liked. When she got it home, she
discovered that it went perfectly with her favorite dress.

Attempt to answer each of the following questions with each program:

Q1: What did Mary go shopping for?
Q2: What did Mary find that she liked?
Q3: Did Mary buy anything?
Program 1

This program is to attempt to answer questions using the literal text. It simply matches text fragments in
the questions against the input text.

Data structures

Question patterns: A set of templates that match common question forms and produce patterns to be
used to match against inputs. Templates and patterns are paired so that if a template matches successfully
against an input question then its associated text patterns are used to try finding appropriate answers in
the text. For example, if the template “who did x y” matches an input question, then the text patterns “x,
y, z” is matched against the text and the value of z is given as the answer to the question.

Text The input text stored simply as a long character string.

Question The current question also stored as a character string.

Algorithm
To answer a question, do the following:

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 11

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
1. Compare each element of Question Patterns, the question and all those that match successfully
to generate a set of text patterns.
2. Pass each of these patterns through a substitution process that generates alternative forms of
verbs so that, for example, “go” in a question might match “went” in the text. This step generates
a new, expanded set of text patterns.
3. Apply each of these text patterns to text, and collect all the resulting answers.
4. Reply with the set of answers just collected.

Answers:

Q1: The template “what did x y” matches this question and generates the text pattern “Mary go
shopping for z.” After the pattern-substitution step, this pattern is expanded to a set of patterns including
“Mary goes shopping for z,” and “Mary went shopping for z.” the latter pattern matches the input
text; the program, using a convention that variables match the longest possible string up to a sentence
delimiter (such as a period), assigns z the value, “a new coat,” which is given as the answer.

Q2: Unless the template set is very large, allowing for the insertion of the object of “find” between it and
the modifying phrase “that she liked,” the insertion of the word “really” in the text, and the substitution
of “she” for “Mary,” this question is hot answerable. If all of these variations are accounted for and the
question can be answered, then the response is “a red one”.

Q3: Since no answer to this question is contained in the text, no answer will be found.

Program 2

This program first converts the input text into a structured internal from that attempt to capture the
meaning of the sentences. It also converts questions into that form. It finds answers by matching
structured forms against each other.

Data structures

EnglishKnow: A description of the words, grammar, and appropriate semantic interpretations of a large
enough subset of English to account for the input texts that the system will see. This knowledge of
English is used both to map input sentences into an internal, meaning-oriented form and to map from
such internal forms back into English. The former process is used when English text is being read; the
latter is used to generate English answers from the meaning-oriented from that constitutes the program’s
knowledge base.

Input text The input text in character form.

Structured Text A Structured representation of the content of the input text. This Structure attempts
to capture the essential Knowledge contained in the text, independently of the exact way that the
Knowledge was stated in English.

A Structured Representation of a sentence:

Event 2

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 12

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Instance: Finding
Tense: Past
Agent: Mary
Object: Thing 1
Thing 1
Instance: Coat
Color: Red
Event 2
Instance: Liking
Tense: Past
Modifier: Much
Object: Thing 1

InputQuestion The input question in character form.

StructQuestions A structured representation of the content of the user’s question. The structure
is the same as the one used to represent the content of the input text.
Algorithm:

Convert the input text into structured form using the Knowledge contained in English Know. This may
require considering several different potential structures, for a variety of reasons, including the fact that
English words can be ambiguous, English grammatical structures can be ambiguous, and pronouns may
have several possible antecedents. Then, to answer a question, do the following:

1. Convert the question to structured form, again using the Knowledge contained in English Know. Use
some special marker in the structure to indicate the part of the structure that should be returned as the
answer. This marker will often correspond to the occurrence of a question word (like “who” or
“what”) in the sentence. The exact way in which this marking gets done depends on the form chosen
for representing structured Text.
2. Match this structured form against Structured Text.
3. Return as the answer those parts of the text that match the requested segment of the question.

Answers:
Q1: This question is answered straightforwardly with, “a new coat”
Q2: This one also is answered successfully with, “a red coat”.
Q3: This one, though, cannot be answered, since there is no direct response to it in the text.

Program 3

This program converts the input text into a structured form that contains the meanings of the
sentences in the text, and then it combines that form with other structured forms that describe prior
knowledge about the objects and situations involved in the text. It answers questions using this
augmented knowledge structure.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 13

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Data structures

World Model: A Structured representation of background world Knowledge. This structure contains
Knowledge about objects, actions and situations that are described in the input text. This Structure is
used to construct Integrated Text from the input text. For example, Figure below shows an example
of a Structure that represents the systems Knowledge about shopping. In the case of this text, for
example, M is a coat and M’ is a red coat. Branches in the figure describe alternative paths through
the script.

The Algorithm:
Convert the input Text into structured form using both the Knowledge contained in English Know
and that contained in world model. The number possible structures will usually be greater now than it
was in program 2 because so much more Knowledge is being used. Sometimes, though, it may be
possible to consider fewer possibilities by using the additional knowledge to filter the alternatives.

Shopping script:
Roles: C (customer), S (salesperson)
Props: M (merchandise), D (dollars)
Location: L (a store)
To answer a question, do the following:

1. Convert the question to structured form as in program 2 but use WorldModel necessary to resolve
any ambiguities that may arise.
2. Match this structured form against Integrated Text.
3. Return as the answer those parts of the text that match the requested segment of the question.

A Shopping Script diagram

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 14

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Answers:
Q1: Same as Program 2.
Q2: Same as Program 2.
Q3: Now this question can be answered. The shopping script is instantiated for this text, and because of
the last sentence, the path through step 14 of the script is the one that is used in forming the
representation of the text. When the script is instantiated M1 is bound to the structure representing the red
coat. After the script has been instantiated, Integrated Text contains several events. That are taken from
the script but that are not described in the original text, including the event “Mary buys a red coat” (from
step 10 of the script). Thus, using the integrated text as the basis for question answering allows the
program to respond “she bought a red coat”.

We can conclude that these problems illustrate important AI techniques:

✓ Search- Provides a way of solving problems for which no more direct approach is available.
✓ Use of Knowledge- Provides a way of solving complex problems by exploiting the structures of the
objects that are involved.
✓ Abstractions – Provides a way of separating important features and variations from the many
unimportant ones.

1.4 THE LEVELS OF THE MODEL / (STATE OF THE ART)

We must ask ourselves, what is our goal in trying to produce programs that do the intelligent things
that people do? Or, are we trying to produce programs that do the tasks the same way people do?
Or, are we attempting to produce programs that simply do the tasks in whatever way appear
easiest? There have been AI projects motivated by each of these goals.

These programs are divided into two classes –

✓ Programs in the first class attempt to solve problems that do not really fit definition of an AI task.
They are problems that a computer cloud easily solves.
✓ Example is Elementary Perceiver and Memorizer (EPAM) [Feigenbaum, 1963], which
memorized associated pairs of nonsense syllables. Memorizing pairs of nonsense syllables is easy
for a computer. But this task is hard for people.
✓ The programs in the second class attempt to model human performance and are within AI
definitions. Reasons for this are:

→ To test psychological theories of human performance: An example of a program that was

written for this reason is PARRY [Colby, 1975], which exploited a model of human paranoid
behavior to simulate the conversational behavior of a paranoid (Suspicious) person. The model
was good enough that when several psychologists were given the opportunity to converse with
the program via a terminal, they diagnosed its behavior as paranoid.
→ To enable computers to understand human reasoning: For example, for a computer to be
able to read a newspaper story and then answer a question, such as “why did the terrorists kill
the hostages?” Its program must be able to simulate the reasoning processes of people.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 15

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
→ To enable people to understand computer reasoning: In many circumstances, people are
reluctant to rely on the output of a computer unless they can understand how the machine
arrived at its result. If the computer’s reasoning process is similar to that of people, then
producing an acceptable explanation is much easier.
→ To exploit what knowledge we can glean (Gather) from people: Since people are the best-
known performers of most of the tasks with which we are dealing, it makes a lot of sense to
look to them for clues as to how to produce.

• The following are the disciplines that contributed ideas, new points, and techniques to AI:
✓ Philosophy: (428BC- Present)
1. Can formal rules be used to draw valid conclusions?
2. How does the mental mind arise from a physical brain?
3. Where does knowledge come from?
4. How does knowledge lead to action?
✓ Mathematics
1. What are the formal rules to draw valid conclusions?
2. What can be computed?
3. How do we reason with uncertain information?
✓ Economics
1. How should we make decisions so as to maximize pay off?
2. How should we do this when others may not go along?
3. How should we do this when the pay off may be far in the feature?
✓ Neuroscience
How do brains process information?
✓ Psychology
How do humans and animals think & act?
✓ Computer Engineer
How can we an efficient as computer?
✓ Computer Theory and Cybernetics
How can computer operate under their own control?
✓ Linguistic
How does language relate to thought?

Note: The questions under each of the above foundations, shows that, what exactly their contributions

• The History / Early works of AI:

✓ The first work in AI was done by McCalloch and Pitts (1943), who proposed artificial neurons.
They worked on three sources:
1. Knowledge of the basic Physiology.
2. Functions of neurons in the brain.
3. A formal analysis of propositional logic and turing’s theory of computations.
✓ Later in 1949, Donald Hebb demonstrated a simple updating rule for modifying the
connection strengths between neurons, called Hebbbian Learning.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 16

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
✓ Alan Turing (1950), was the first to articulate the complete vision of AI in his article (1950) –
“Computing Machinery and Intelligence”, he introduced the Turing Test, Machine
learning, genetic algorithms and reinforcement learning.
✓ John McCarthy of Dart Month College called the father of AI, introduced automata theory,
neural networks, and the study of intelligence (1956). Also he designed the language LISP in
1958.
✓ Newell and Simon (1976) designed a program called General Problem Solver (GPS) that
imitated human problem solving protocols. GPS was the first program to embody the
“Thinking humanly” approach.
✓ Rosenblatt (1962) developed perceptrons by enhancing the Hebban learning algorithm.
✓ DENDRAL and MYCIN were the two expert systems – one for Molecule structure analysis
and the other for diagnose the blood infections, developed during 1969, at MIT.
✓ Recent years have seen a revolution in both the content and the methodology of work in AI.
✓ Complete Agent architecture proposed during 1990 aims to understand the workings of
agents embedded real environments, with continuous sensory inputs. One of the most
important environments for intelligent agents in the Internet.
✓ AI Systems have become so common in web-based applications including search engines,
recommended systems, and website construction systems.

• What can AI do today?

These are just examples of AI systems that exist today.
✓ Autonomous Planning and scheduling: NASA’S Remote Agent program became the first
on board autonomous planning program to control the scheduling of operations for a
spacecraft (Jonsson et.al., 2000), such as detection, diagnosis and recovery from problems as
that occurred.
✓ Game Planning: IBM’S Deep Blue, Became the first computer to defect the world
champion in chess match when it beated Garry Karpasov (1997), the value of IBM’S stock
increased by $18 billion.
✓ Autonomous Control: The ALVINN computer vision system was trained to steer a car to
keep it following a lane (driving autonomously 98% of the time from Pittsburgh to San
Diego, 2850 miles)
✓ Diagnosis: Medical diagnosis programs based on probabilistic analysis have been able to
perform at the level of an expert physician is several areas of medicine.
✓ Logistics Planning: During the Persian Gulf crisis of 1991, US forces deployed a Dynamic
Analysis and Re-planning Tool, (DART) in1994, to do automated logistics planning and
scheduling for transportation. This involved up to 50,000 vehicles, cargo, and people at a
time, and had to account for starting points, destinations, routes and conflict resolution among
all parameters.
✓ Robotics: Many surgeries now use robot assistants in microsurgery. Ex: HIPNAV (1996) is a
system that uses computer vision techniques to create a three- dimensional model of a
patient’s internal anatomy and the uses robotic control to guide the insertion of a hip
replacement prosthesis.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 17

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
✓ Language Understanding and Problem Solving: PROVERB (1999) is a program that
solves crossword puzzles setter than most humans.

1.5 CRITERIA FOR SUCCESS

✓ One of the most important questions to answer in any scientific or engineering research project is
“How will we know if we have succeeded?” Artificial intelligence is no exception. How will we
know if we have constructed a machine that is intelligent? Can we do anything to measure our
progress? Yes
✓ In 1950, Alan Turing proposed the following method for determining whether a machine can
think. His method has since become known as the Turing Test. To conduct this test, we need
two people and the machine to be evaluated. One person plays the role of the interrogator, who
is in a separate room from the computer and the other person. The interrogator can ask questions
of either the person or the computer by typing questions and receiving typed responses. However,
the interrogator knows them only as A and B and aims to determine which the person is and
which the machine is. The goal of the machines is to fool the interrogator into believing that it is
the person. If the machine succeeds at this, then we will conclude that the machine can think.
The machine is allowed to do whatever it can to fool the interrogator. So, for example, if asked
the questions “How much is 12,324 times 73,981?” it could wait several minutes and then
respond with the wrong answer [Turing, 1963].
✓ We are forced to conclude that the question of whether a machine has intelligence or can think
is too vague to answer precisely. But it is often possible to construct a computer program that
meets some performance standard for a particular task. That does not mean that the program does
the task in the best possible way. It means only that we understand at least one way of doing at
least part of a task.

1.6 SOME GENERAL REFERENCES

There is a great many sources of information about artificial intelligence.

✓ First, will have some survey books: The broadest are the multi-volume handbook of artificial
intelligence [Barr et al.. 1981] and Encyclopedia of artificial intelligence [Shapiro and Eckroth,
1987], both of which contain articles on each of the major topics in the field.
✓ Four other books that provide good overviews of the field are artificial intelligence [Winston,
1984], introduction to artificial intelligence [Charniak and McDermott, 1985], Logical
Foundations of artificial intelligence [Genesereth and Nilsson, 1987], and The Elements of
artificial intelligence [Tanimoto, 1987] of more restricted scope is principles of artificial
intelligence [Nilsson, 1980], which contains a formal treatment of some general- purpose AI
techniques.
✓ Most the work conducted in AI has been originally reported in journal articles, conference
proceedings or technical reports. But some of the most interesting of these papers have later
appeared in special collections published as books. Computer and Thought [Feigenbaum and
Feldman, 1963] is a very early collection of this sort. Later ones include Simon and
Siklossy[1972], Schank and Colby [1973], Bobrow and Collins [1975], waterman and Hayes-
Roth[1978], Findler [1979],Webber and Nilsson [1981], Halpern[1986], Shrobe [1988], and

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 18

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
several others that are mentioned in later chapter in connection with specific topics. For newer AI
paradigms the book fundamentals of the new artificial intelligence [Toshinori Munakata, 1998] is
a good one.
✓ The major journal of AI research is called simply Artificial Intelligence. In addition, Cognitive
science is devoted to papers dealing with the overlapping areas of psychology, linguistics, and
artificial intelligence. AI magazine is a more ephemeral, less technical magazine that is published
by the American Association for artificial intelligence (AAAI). IEEE Expert, IEEE Transactions
on Systems, Man and Cybernetics, IEEE Transactions on Neural Networks and several other
journals publish papers on a board spectrum of AI application domains.
✓ Since 1969, there has been a major AI conference, the International Joint Conference on Artificial
Intelligence (IJCAI), held every two years. The proceedings of these conferences give a good
picture of the work that was taking place at the time. The other important AI conference, held
three out of every four years starting in 1980, is sponsored by the AAAI, and its proceedings, too,
are published.
✓ In addition of these general references, there exists a whole array of papers and books describing
individual AI projects.

1.7 ONE FINAL WORD AND BEYOND

Solving the problems is the topic of discussion in AI. We need methods to help us solve AI’s serious
dilemma:
✓ An AI system must contain a lot of knowledge if it is to handle anything
✓ But as the amount of knowledge grows, it becomes harder to access the appropriate things when
needed, so more knowledge must be added to help. But now there is even more knowledge to
manage, so more must be added, and so forth.
✓ AI is still young discipline possibly in the sense that little has been achieved as compared to what
was expected.
✓ Robots form the ultimate test-bed for AI. Finally one should not forget that research in AI is
multidisciplinary.

Chapter 2: Problems, Problem Spaces, and Search

To build a system to solve a particular problem, we need to do four things:

1. Define the problem precisely: This definition must include precise specifications of what the
initial situation (s) will be as well as what final situations constitute acceptable solutions to the
problem.
2. Analyze the problem: A few very important features can have an immense impact on the
appropriateness of various possible techniques for solving the problem.
3. Isolate and represent the task knowledge that is necessary to solve the problem.
4. Choose the best problem-solving technique(s) and apply it (them) to the particular problem.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 19

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
1.8 DEFINING THE PROBLEM AS A STATE SPACE SEARCH
• AI Problem Solving
✓ Formal Description of a Problem
Identity the objects /entities involved in the problem.
Specify the Initial State (status) of the objects. i.e., the initial configurations / situations
from where the problem solving process starts.
Specify the goal state that would be acceptable as solution to the problem. i.e, one or more
states / configurations acceptable as solution states.
Define the state space that contains all possible states / configurations / situations of
the objects either in the form of tree or graph.
Specify the set of rules/productions that describe available actions (operators) to get
from one state to another.
Define search Techniques that systematically consider all possible action sequences to
find a path from the initial to goal state.

Also think of: what knowledge is given (restrictions / constraints), knowledge to be assumed,
additional assumptions to be made and order of applying rules (control strategy).

Note: AI Problem Solving → Searching for a goal state.

Example Problems:

1. Playing Chess
✓ Each position can be described by an 8x8 array.
✓ Initial Position is the game opening position.
✓ Goal Position is any position in which the opponent does not have a legal move and his/her
king is under attack, (check made).
✓ Legal Moves can be described by a set of rules.
✓ Left side of the rule matched against the current state.
✓ Right side of the rule describes the new resulting state.
✓ State space is the set of legal positions.
✓ Starting at the initial state, using the set of rules to move from one state to another and
attempting to end up in a goal state is the procedure for solving problem.

One Legal Chess Move

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 20

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
2. Water- Jug Problem
“You are given two jugs, a 4-litre one and a 3-litre one. Neither have any measuring markers on
it. There is a pump that can be used to fill the jugs with water. How can you get exacting 2 liters
of water into 4- litre jug?”

Solution: objects water jugs (2) pump let ‘x’ represents amount of water in 4-litre jug ‘y’ 3 litre
jug.
➔ State represents pair (x,y) where x = 0, 1, 2, 3, or 4
y = 0, 1, 2, or 3

x- Amount of water in jug 1(4-litre)

y- jug 2 (3-litre)

➔ Initial / starting state: (0,0)

➔ Goal / final state: (2, n) for any ‘n’
➔ Rules / productions / operators

1. (x,y) / if x < 4 → (4, y) Fill jug 1

2. (x,y) / if y < 3 → (x,3) Fill jug 2
3. (x,y) / if x > 0 → (x-d, y) Pour some water out of jug 1
4. (x,y) / if y > 0 → (x, y-d) Pour some water out of jug 2
5. (x,y) / if x > 0 → (0, y) Empty jug 1 on ground
6. (x,y) / if y > 0 → (x, 0) Empty jug 2 on ground
7. (x,y) / if y >0 and x+y ≥ 4 → (4, y-(4-x)) Pour some water from jug 2 to jug 1 till
jug 1 is full
8. (x,y) / if x > 0 and x+y ≥ 3 → (x-(3-y), 3)) Pour some water from jug 1 to jug 2 till
jug 2 is full
9. (x,y) / if y > 0 && x+y ≤ 4 → (x+y, 0) Pour complete water from jug 2 to jug 1
10. (x,y) / if x > 0 && x+y ≤ 3 → (0, x+y) Pour complete water from jug 1 to jug 2
11. (0,2) → (2, 0) Special rules to capture knowledge at
12. (2,y) → (0, y) some stage in solving a problem

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 21

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)

➔ Solution Space Tree / State Space Tree

Note: Starting from (0, 0), search and loop until it reaches the goal state (2, 0). Apply a rule whose left
side matches the current state and set the new current state to be the resulting state.

Similarly, try for 6-litre and 8-litre jugs; indicate how 8-litre jug can be filled with 4 litre of water.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 22

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
→ (0,0)
→ (0,8) separate all the rules & state space tree (First Possible solution)
→ (6,2)
→ (0,2)
→ (2,0)
→ (2,8)
→ (6,4)

→ (0,0)
→ (6,0) separate all the rules & state space tree (Second Possible solution)
→ (0,6)
→ (6,6)
→ (4,8)
→ (4,0)
→ (0,4)

Additional Programs
3. Man –Tiger –Cow - Grass Problem [MTCG]

A man has a pet Tiger, a cow, and a bundle of grass with him. He wants the cross from one Bank (A) to
another Bank (B) of a river using a small boat that can hold him and any one of the remaining three. If
the man is not there, Tiger would eat cow, and similarly, the cow would eat the grass. Man does not want
to lose any of his possessions. Indicate how he can cross the river (BANK A to BANK B).

➢ State space approach for solving the MTCG problem.

Initial State: Man (M), Tiger (T), Cow (C), Gross (G) on BANK A and none on BANK B
Goal State: ø; MTCG on BANK B
Action / Operation / rule:

Move (object, Source, Destination)

Bank A Bank B Bank A Bank B

T/C/G/none

Meaning: Man transports the object in the boat from source to destination.

✓ Knowledge Base (KB)

1. The Tiger eats the cow if man is not with them.
2. The Cow cats the grass if man is not with them.
3. The Tiger does not eat the Grass
4. The Tiger does not eat the man. (assumed)
5. Boat carry only two objects (Man & one of his pets)
6. Man has to ride the boat.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 23

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)

✓ Actions and Conditions

Rule No Action Conditions

1 Move(T, Bank-A, Bank-B) T is on Bank-A with C or G
2 Move(C, Bank-A, Bank-B) C is on Bank-A with T/G/ both / ø
3 Move(G, Bank-A, Bank-B) G is on Bank-A with T/C
4 Move(T, Bank-B, Bank-A) T is on Bank-B with C/G
5 Move(C, Bank-B, Bank-A) C is on Bank-B with T/G
6 Move(G, Bank-B, Bank-A) G is on Bank-B with T/C

✓ If - then rules for MTCG problem

1. If state S= (MTC; G) / (MTG; C) then move (T, Bank-A, Bank-B)
2. If S=(MC; TG) / (MTCG; ø) / (MCG;T) / (MCT;G) then move (C, Bank-A, Bank-B)
3. If S=(MGC; T) / (MGT; C) then move (G, Bank-A, Bank-B)
4. If S=(G; MTC) / (C; MTG) then move (T, Bank-B, Bank-A)
5. If S= (T; MCG) / (G; MCT) then move (C, Bank-B, Bank-A)
6. If S= (T; MGC) /(C; MGT) then move (G, Bank-B, Bank-A)
7. If S=(TG; MC) / (C; MTG) then move (∅, Bank-B, Bank-A)

✓ One Possible Solution for MTCG problem using state space approach

State after applying the

State No. Operator/Action
Operator
1 (MTCG: ∅ )
2 Move (C, Bank-A, Bank-B) (TG : MC)
3 Move (∅, Bank-B, Bank-A) (MTG: C)
4 Move (G, Bank-A, Bank-B) (T: MCG)
5 Move (C, Bank-B, Bank-A) MTC: G)
6 Move (T, Bank-A, Bank-B) C: MTG)
7 Move (∅, Bank-B, Bank-A) MC: TG)
8 Move (C, Bank-A, Bank-B) (∅: MTCG) goal state

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 24

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)

4. Three – Married Couple Problem

Three married couples come to the river bank A. there is a boat on bank A that can carry at most two
persons. All the six persons want to cross the river and reach bank B using the boat, but there are certain
condition / restrictions to be observed and met.

i) No lady wants to row the boat

ii)No lady wants travel in the boat with a man other than her husband
iii)
No lady wants to be left alone on a bank.
iv)No lady wants to stay on a bank with other man when her husband is not there, unless she has
the company of other lady.
v) No man wants to row the boat more than thrice (3).
a) Give a possible solution to this problem.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 25

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
b) Write state- space representation of the above problem indicating different states, operators, and
conditions for using the operators.

Solution:

✓ Objects
1. 1 to 3 - married couples C1, C2, C3

m1f1, m2f2, m3f3 six persons

C1 C2 C3
1 boat

✓ Knowledge Base
1. Boat can carry at most 2 persons
2. All the conditions / restrictions mentioned above.

Initial state: m1f1 m2f2 m3f3; ø

Goal State: ø; m1f1 m2f2 m3f3
Operator: MOVE (Persons, Source, Destination)
Meaning that male person (husband) row the boat & move a zero or one male / female
person from source to destination.

Source: Bank-A or Bank-B

Destination: Bank-A or Bank-B.
✓ One Possible Solution

State No Operator /Action State after applying operator

Bank A Bank B
∅
1 Move Src Destn m1m2m3 f1f2f3
initial state
2 {1} m1 m2 A B m3 f1 f2 f3 m1 m2
3 {2} m1 B A m1m3f1f2f3 m2
4 {3} m1 f1 A B m3 f2 f3 m1 m2 f1
5 {1} m2 B A m2 m3 f2 f3 m1f1
6 {2} m2 f2 A B m3 f3 m1 f1 m2 f2
7 {3} m2 B A m2 m3 f3 m1 f1 f2
8 {1} m3 f3 A B m2 m1 f1 f2 m3 f3
9 {2} m3 B A m2 m3 m1 f1 f2 f3
m1 f1 m2 f2 m3 f3
10 {3} m2 m3 A B ∅
goal state

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 26

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)

5. Missionaries and Cannibals Problem

Three Missionaries and three cannibals want to cross a river. There is a boat on their side of the river that
can be used by either one or two persons. How they cross the river?

Note: at any point of time, on either side of the river, number of cannibals should never be more than
missionaries. How can they all cross over without anyone being eaten?

Solution:

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 27

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
✓ Objects: 3 missionaries, 3 cannibals, 1 boat

Bank-A Bank-B

Start State: 3M3C: ø

Goal State: ø: 3M3C

✓ Operator /rule: move (persons, source, destination)

Persons: missionaries or cannibals
Source: Bank-A / Bank-B
Destination: Bank-A / Bank-B

✓ Knowledge given
1. Boat can carry at most two persons.
2. Any Person can row the boat
3. Number of cannibals should never be more than number of missionaries at any point of time on
either side of the river.

✓ One Possible Solution

State No. Operator move Bank-A Bank-B

∅
1 3M3C
initial State
2C
2 3M1C 2C
⟶
1C
3 1C
⟵ 3M2C
2C
4 3M 3C
⟶
1C
5 3M1C 2C
⟵
2M
6 1M1C 2M2C
⟶
1M1C
7 2M2C 1M1C
⟵
2M
8 2C 3M1C
⟶
1C
9 3C 3M
⟵
2C
10 1C 3M2C
⟶
1C
11 2C 3M1C
⟵
2C 3M3C
12
⟶ ∅ goal state

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 28

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
✓ State Space representation for Missionaries and Cannibals Problems:

6. The Monkey and Bananas Problem

A hungry monkey finds himself in a room in which a bunch of bananas is hanging from the ceiling. The
monkey, unfortunately cannot reach the bananas. However, in the room there are also a chair and a stick.
The ceiling is just the right height so that a monkey standing on a chair could knock the bananas down

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 29

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
with the stick. The monkey knows how to make around, carry other things around, reach for the bananas,
and wave a stick in the air. What is the best sequence of actions for the monkey to take to acquire lunch?

Solution:
Actions in sequence are:
1. The monkey reaches to the box.
2. Pushes the box under hanging bananas
3. Climbs on the box.

The state of the problem consists of representing the states by four elements (X, W, Y, and K)
Where X= horizontal position of monkey
W= Monkey’s position on the box, if yes w=1, not w=0
Y= horizontal position of box
K= grasp of bananas by monkey, if yes k=1, not k=0
Four operators govern the four possible actions
Go to (P): Monkey goes to horizontal position P
Push box (Q): Monkey pushes the box to horizontal position Q
Climb box ( ): Monkey climbs top of the box
Graph ( ): Monkey holds the bananas
P and Q are the variables to represent any position in the room for the monkey and box respectively.
State Action Result

(x, 0, y, k) Go to (P) (p, 0, y, k)

Money &box are at X becomes available P

Different positions

(x, 0, x, k) push box (a) (q, 0, q, k)

Monkey & box are monkey and box have the same
at same position common position variable q?.

(x, 0, x, z) Clims box (c) (x, 1, x, k)

Common static position x a common static position

When monkey is on box.
For monkey &banana
(z, 1, z, 0) group (z, 1, z, 1)
Z is the location on the floor the monkey on the box grasps banana.
directly under the banana

7. Eight- Puzzle Problem

The 8-puzzle problem consists of a 3x3 grid, which holds 8 movable tiles numbered as 1 to 8. One
square is empty, allowing the adjacent tiles to be shifted. The objective of the puzzle is to find a sequence

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 30

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
of tile movements that leads from a starting configuration to a goal configuration. As shown in below
example.

To be transformed
1 2 3 1 2 3
8 5 6
⟹ 4 5 6
4 7 7 8
Initial state / configuration (start) Goal state/configuration (Final)

Note: i) States of 8-puzzle are the different permutations of the tiles within the frame.
ii) An optimal solution is the one that maps an initial arrangement of tiles to the goal state
with the smallest number of moves.
iii) Root node can be any random starting state.

Solution:

States: Specifies the locations of each of the tiles and the blank in one of the nine squares.

Initial State: Any state can be designated as initial state.

Goal state: Many goal configurations are possible.

Legal moves: Operators ⟹

1. Move blank tile left (L)

2. Move blank tile right (R)
3. Move blank tile Up (U)
4. Move blank tile Down (D)

Search Space: A tree that shows possible moves from an initial state.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 31

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 32

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)

8. The Travelling Salesman Problem (TSP)

A salesman has a list of cities, each of which he has to visit exactly once. There are roads between each
pair of cities on the list. i.e., n-cities with paths connecting the cities.

Find the route (tour) the salesman should follow for the shortest possible round trip that starts and ends at
any one of the cities.

Diagrammatically,

5
A B
2

2 3 1 4 E

3
C D
6

5 Cities: A, B, C, D, E and cost / distance is given.

The objective of TSP is to find a tour with minimum cost/distance.
Typically tour would be: (starting from A) and visits all

A B
2

2 1 3
E
C 3

Total Cost=2+1+2+3+3=11

Note:

i) To explore all such tours requires an exponential amount of time.

ii) As long as ‘N’ is small this approach works but it breaks down as numbers of cities grow.
iii) If there are ‘N’ cities, then the number of different path among them is 1.2… (N-1), or (N-1)!

The time to examine single path is proportional to N. So, the total time required to perform this search
is proportional to N!
For ex: N=10, means 10! i.e., 3,628,800 time units (large number)

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 33

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)

1.9 PRODUCTION SYSTEMS (Rule based systems) or (Inferential systems)

✓ Since search forms the core of many intelligent process, it is useful to structure AI programs in a
way that facilities describing the search process. So, production systems provide the structure for
AI programs that makes searching process easy. i.e., production systems help in describing and
performing the search operation in AI programs.

✓ Production System Consists of

A set of rules of the form x → y
LHS = Patterns and RHS = Action
LHS determines applicability of rule, RHS specifies action.
Database OR knowledge base of the domain.
Control Strategy: That specifies the order in which rules are applied on the knowledge base
and derive the new knowledge.
✓ In general, the process of solving a problem can usefully be modeled as a production system. i.e,
any computable procedure can be modeled as production system.

✓ Advantages
• It models strong data-driven nature of intelligent action
• New rules can easily be added to take care of new situations without disturbing the rest of the
system
• Handling of changes in knowledge base is dynamic
• Helps in making inference mechanism easy.

➔ PRODUCTION SYSTEM CHARACTERISTICS

✓ 4 categories of Production Systems

Monotonic Non-Monotonic
Partially Ex: theorem Ex: Blocks world, 8-
commutative proving puzzle
Non-
Ex: chemical Ex: Bridge/cards
partially
Synthesis problem
commutative
✓ A Monotonic Production System: is one in which the application of a rule never prevents
the later application of another rule that could also have been applied at the time the first
rule was selected.
✓ A Non-Monotonic Production System: is a production system is one in which the above
statement is not true.
✓ A partially commutative production system: is a production system with the property
that if the application of a particulars sequence of rules transforms state ‘X’ into state ‘Y’
then any permutation of those rules that is allowable also transforms state ‘X’ into state
‘Y’. Partially commutative monotonic production systems are useful for solving

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 34

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
ignorable problems and these can be easily implemented without the ability to backtrack
to previous states when it is discovered that an incorrect path has been followed.
✓ Non- Partially commutative production system: are useful for many problems in which
irreversible changes occur.
✓ Non-Monotonic Partially commutative systems: are useful for problems in which
changes occur but can be reversed in which order of operations is not critical.
✓ Commutative production system is one which is both monotonic and partially
commutative.

➔ Control Strategies
✓ Specifies what rule to be applied next and in what order / sequence during the search
process. There are many control Strategies exist uniformed search strategies like DFS,
BFS and informed / heuristic Strategies like Hill climbing, Backtracking, Branch and
Bound, Best first search etc.,.
✓ Two requirements for good control strategies are:
• They should cause motion, so that it lead to a solution
• They should be systematic, so that it lead to a solution
Ex: BFS and DFS strategies

Algorithm: Breadth-First Search

1. Create a variable called NODE-LIST and set it to the initial state.
2. Until a goal state is found or NODE-LIST is empty:
a) Remove the first element from NODE-LIST and call it E. If NODE-LIST was empty,
quit.
b) For each way that each rule can match the state described in E do.
i) Apply the rule to generate a new state,
ii) If the new state is a goal state, quit and return this state.
iii) Otherwise, add the new state to the end of NODE-LIST.

Algorithm: Depth-First Search

1. If the initial state is a goal state, quit and return success.
2. Otherwise, do the following until success or failure is signaled:
a) Generate a successor, E, of the initial state. If there are no more successors, signal
failure.
b) Call depth-First Search with E as the initial state.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 35

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
c) If success is returned, signal success. Otherwise continue in this loop.

Advantages of Depth-First Search

✓ Depth-first Search requires less memory since only the nodes on the current path are stored.
This contract with breadth-first search, where all of the tree that has so far generated must be
stored- more Memory.
✓ Depth-first search may find a solution without examining much of the search space at all.
This contracts with breadth-first search in which all parts of the tree must be examined to
level n before any nodes on level n+1 can be examined.

Advantages of Breadth-First Search

✓ Breadth-first search will not get trapped exploring a blind alley (path). This contrast with
depth-first searching, which may follow a single unfruitful path for a very long time, before
the path actually terminates in a state that has no successors. This is a particular problem in
depth-first search if there are loops (i.e., a state has a successor that is also one of its
ancestors) unless special care is expended to test such a situation.
✓ If there is a solution, then breadth-first search is guaranteed to find it. Furthermore, if there
are multiple solutions, then a minimal solution i.e., one that requires the minimum number
of steps will be found. This is guaranteed by the fact that longer paths are never explored until
all shorter ones have already been examined. These contrasts with depth-first search, which
may find a long path to a solution in one part of the tree, when a shorter path exists in some
other, unexplored part of the tree.

The Travelling Salesman Problem

A Salesman has a list of cities, each of which he must visit exactly once. There are direct
roads between each pair of cities on the list. Find the route the Salesman should follow for the
shortest possible round trip that both starts and finishes at any one of the cities.

A simple, motion-causing and systematic control structure could, in principle, solve this
problem. It would simply explore all possible paths in the tree and return the one with the
shortest length. This approach will work in practice for very short lists of cities. But it breaks
down quickly as the number of cities grows. If there are N cities, then the number of
different paths among them is 1.2… (N-1), or (N-1)! . The time to examine a single path is
proportional to N. So the total time required to perform this search is proportional to N! .
Assuming there are only 10 cities, 10! is 3,628,800, which is a very large number. The

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 36

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
salesman could easily have 25 cities to visit. To solve this problem would take more time than
he would be willing to spend. This phenomenon is called combinatorial explosion. To
combat it, we need a new control strategy.

➔ Heuristic Search

✓ A Heuristic function is a function that maps from problem state description to measures of
desirability, usually represented as numbers.
✓ Improves the efficiency of a search process and always finds a very goal solution for hard
problems.
✓ In general, it points in interesting directions.
✓ Like tour guides, helps to guide a search process. Ex: nearest Neighbor Heuristic
✓ Well designed Heuristic functions can play an important part in efficiently guiding a search
process toward a solution.
✓ Sometimes very simple Heuristic functions can provide a fairly good estimate of whether a
path is good or not. In other situations, more complex Heuristic functions should be
employed.
Note:

1. Production systems represent both knowledge as well as action.

2. Production systems provide a language in which the representation of expert knowledge is very
natural.
3. Production systems provide a Heuristic model for human behavior.
4. A good solution to a problem requires an efficient control strategy.

1.10 AI PROBLEM CHARACTERISTICS

In order to choose the most appropriate method for a particular problem, it is necessary to analyze the
problem along several key dimensions / Characteristics:

1. Is the problem Decomposable into a set of independent smaller and/or easier sub problem? (D)
2. Can solution steps be Ignored or at least undone if they prove unwise? (I)
3. Is the problem’s universe Predictable? [predictability (P)]
4. Is a good solution to the problem obvious without comparison to all other possible solutions?
[Comparability ( C )]
5. Is the desired solution a state of the world OR a path to a state? (R)
6. Is a large amount of knowledge absolutely required to solve the problem, OR is knowledge
important only to constrain the search? [knowledge based consistency (C )]
7. Can a computer that is simply given the problem return the solution, or will the solution of the
problem require interaction between the computer and a person? [ Interactions (I) ]

Note: Above characteristics may be abbreviated as D I P C R C I for easy remembrance.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 37

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
1. Is the problem decomposable?
Suppose the following expression is given, then it could be solved as follows. It can be broken
down or decomposed into three smaller problems, each of which can be solved using small
collection of specific rules.

Now consider a Simple Blocks World Problem illustrated below. This problem is drawn from the
domain often referred to in AI literature as the blocks world. Assume that the following operators
are available:
1. CLEAR (x) [block x has nothing on it] → ON (x Table) [pick up x and put it on the
table]
2. CLEAR (x) and CLEAR (y) → ON (x, y) [put x on y]

Start: Goal:
A

B
C
C
A B

ON (C, A) ON (B, C) and ON (A, B)

Applying the technique of problem decomposition to this simple blocks world example, would
lead to a solution tree as shown in Figure below. In the figure, goals are underlined. States that
have been achieved are not underlined. The idea of this solution is to reduce the problem of
getting B on C and A on B to two separate problems. The first of these new problems, getting B on
C, is simple, given the start state. Simply put B on C. The second sub goal is not quite so simple.

Figure: A Proposed Solution for the Blocks Problem

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 38

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)

ON (B, C) and ON (A, B)

ON (B, C) ON (A, B)

ON (B, C) CLEAR (A) ON (A, B)

CLEAR (A) ON (A, B)

2. Can solution steps be Ignored or Undone?

These three problems: theorem proving, 8-puzzle and chess illustrate the differences between
three important classes of problems:

Example:
• Ignorable (e.g.., theorem proving), in which solution steps can be ignored
• Recoverable (e.g.., 8-puzzle), in which solution steps can be undone
• Irrecoverable (e.g.., chess), in which solution steps cannot be undone

These three definitions make reference to the steps of the solution to a problem and thus may
appear to characterize particular production system for solving a problem. This is true for each of
the problem used as examples above. When this is the case, it makes sense to view the
recoverability of a problem as equivalent to recoverability of a natural formulation of it.

3. Is the problem’s Universe Predictable?

The planning process can only be done effectively for certain-outcome. A few examples of such
problems are:
✓ Playing bridge: we can do fairly well since we have available accurate estimates of the
probabilities of each of the possible outcomes.
✓ Controlling a robot arm: The outcome is uncertain for a variety of reasons. Someone
might move something into the path of the arm. The gears of the arm might stick. A slight
error could cause the arm to knock over a whole stack of things.
✓ Helping a lawyer decide how to defend his client against a murder charge: Here we
probably cannot even list all the possible outcomes, much less assess their probabilities.
Examples:
In 8-puzzle problem → certain outcome
In bridge problem → uncertain outcome (cards)
In controlling a Robot arm → uncertain outcome

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 39

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
In helping a lawyer decides how to depend his client against a murder charge. (cannot list all
possible outcomes)

4. Is a good solution absolute OR Relative?

Example: In the problem of drawing conclusions from a given set of premises, we may need to
consider the facts in some sequence and try to relate them to find new facts which can intern be
used as premises in conclusion drawing process. So, we will have a relative path from facts /
premises to the conclusion.

Example: Marcus was a man.

Marcus was a Pompeian.
Marcus was born in 40 AD.
All men are mortal.
All Pompeian’s died when the volcano erupted in 79 AD.
No mortal lives longer than 150 years it is now 1991 AD.
Suppose, we ask the question, “Is Marcus alive?” No.
Here, Solution will be a relative path.

Since we are interested in the answer to the question, it does not matter which path we follow. If
we do follow one path successfully to the answer, there is no reason to go back and see if some
other path might also lead to a solution. These types of problems are called as “Any path
Problems”. Now consider the Travelling Salesman Problem. Our goal is to find the shortest path
route that visits each city exactly once

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 40

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)

Suppose we find a path it may not be a solution to the problem. We also try all other paths. The
shortest path (best path) is called as a solution to the problem. These types of problems are known
as “Best path” problems. But path problems are computationally harder than any path problems.

5. Is the solution a state or a path?

Consider the problem of finding a consistent interpretation for the sentence The bank president
ate a dish of pasta salad with the fork There are several components of this sentence, each of
which may have more than one interpretation. Some of the sources of ambiguity in this sentence
are the following: The word “Bank” may refer either to a financed institution or to a side of
river. But only one of these may have a President. The word “dish” is the object of the
word “eat”. It is possible that a dish was eaten. But it is more likely that the pasta salad in
the dish was eaten. Because of the interaction among the interpretations of the constituents
of the sentence some search may be required to find a complete interpreter for the sentence.
But to solve the problem of finding the interpretation we need to produce only the interpretation
itself. No record of the processing by which the interpretation was found is necessary. But with the
“water-jug” problem it is not sufficient to report the final state we have to show the “path” also. So
the solution of natural language understanding problem is a state of the world. And the solution of
“Water jug” problem is a path to a state.

Example: In man, Tiger, Cow, & Grass problem, solution is a state.

In Travelling salesman problem solution is a path.

6. What is the Role of the Knowledge? Is the Knowledge base consistent?

Example: In playing chess, role of knowledge is little-just the rules for legal moves & simple
control mechanisms whereas, in the problem of scanning daily news papers and deciding which
are supporting the democrats and which are supporting republicans in some election, the role of
knowledge is extensive. Knowledge base used for solving a problem should be consistent.

7. Does the task require interactions with a person?

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 41

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Can we compute simply the given problem and return the solution? OR will the solution of a
problem requires user interaction?

For Example:
• Solitary problem, in which there is no intermediate communication and no demand for an
explanation of the reasoning process.
• Conversational problem, in which intermediate communication is to provide either additional
assistance to the computer or additional information to the user.

8. Problem Classification
There are several broad classes into which the problems may fall.
Example: In Diagnostic tasks, i.e. both in medical and fault diagnosis, we need to examine the
input and decide which of a set of known classes, the input is an instance of?

1.11 ISSUES IN THE DESIGN OF SEARCH PROGRAM

Every search process can be viewed as a traversal of a tree structure in which each node represent a
problem state and each arc represents a relationship between the states represented by the nodes it
connects. For example, Figure below shows part of a search tree for a water jug problem. The search
process must find a path or paths through the tree that connect an initial state with one or more
final states. The tree that must be searched could in principle, be constructed in it’s entirely from the
rules that define allowable moves in the problem space.

Some important issues that arise in all of them:

✓ The direction in which to conduct the search (forward versus backward reasoning).
✓ How to select applicable rules (matching). Production systems typically spend most of their
time looking for rules to apply, so it is critical to have efficient procedures for matching rules
against states.
✓ How to represent each node of the search process. For problems like chess, a node can be
fully represented by a simple array. In more complex problem solving, however, it is
inefficient and/or impossible to represent all of the facts in the world and to determine all of the
side effects an action may have.

A Search Tree for the Water Jug Problem

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 42

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
For example, in the tree shown above, the node (4, 3), can be generated either by first filling the 4-gallon
jug and then the 3-gallon one. This example also illustrates another problem that often arises when the
search process operates as a tree walk. On the third level, the node (0, 0) appears. But this is the same as
the top node of the tree, which has already been expanded. Those two paths have not gotten us anywhere,
so we would like to eliminate them and continue only along the other branches. The waste of effort that
arises when the same node is generated more than once can be avoided at the price of additional
bookkeeping. Instead of traversing a search tree, we traverse a directed graph. This graph differs from a
tree in that several paths may come together at a node. The graph corresponding to the tree is shown in
figure below. Any tree search procedure that keeps track of all the nodes that have been generated so
far can be converted to a graph search procedure by modifying the action performed each time a node
is generated. Graph search procedures are useful for dealing with partially commutative production
systems.

A Search Graph for the Water Jug Problem

Chapter 3: Heuristic Search Techniques

Many of the problems that fall within the preview of artificial intelligence are too complex to be solved
by direct techniques. They must be attacked by appropriate search methods armed with whatever direct
techniques are available to guide the search. Here, a frame work for describing search methods is
provided and several general purpose search techniques are discussed. These methods are all varieties of
heuristic search. They can be described independently of any particular task or problem domain.
Some Techniques presented here are:
Generate-and-test Hill climbing Best-first-search
Problem reduction constraint satisfaction Means-ends analysis

1.12 GENERATE-AND-TEST-ALGORITHM
✓ It is a simple and exhaustive search method

Steps:

1. Generate a possible solution and solution is either to find a point or path in solution
space.
2. Test whether it is correct solution or not. If yes, return success and quit.
3. Otherwise, return to step 1

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 43

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Advantages:

✓ Useful for simple problems

✓ Similar to DFS
✓ Efficient for problems having large state space search

Disadvantages:

✓ There will be no guarantee that solution will ever be found

For example, consider the puzzle that consists of four six-sided cubes, with each side of each cube
painted one of four colors. A solution to the puzzle consists of an arrangement of the cubes in a row such
that on all four sides of the row one block face of each color is showing. This problem can be solved by a
person in several minutes by systematically and exhaustively trying all possibilities. It can be solved even
more quickly using a heuristic generate-and-test procedure.

1.13 HILL CLIMBING ALGORITHM

✓ Is a variant of generate-and-test, in which, the feedback from the procedure is used to help the
generator to decide, in which direction to move in the search space.
✓ It is so called because of the way the nodes are selected for expansion. In generate and test, the
test function responds with only yes or no. But if the test function is augmented with a heuristic
function that provides an estimate of how close a given state is to a goal state.
✓ At each point in the search path, a successor node that appears to lead most quickly to the top of
the hill (goal) is selected for exploration.
✓ This is used when good Heuristic function is available for evaluating states but no other useful
knowledge is available.
✓ Can be employed if the goal path does not matter.
✓ It is a local search method on state space landscape.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 44

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
• Simple Hill Climbing

Algorithm Steps:

1. Generate the first proposed solution (initial state). Evaluate the initial state. If it is a solution state
then return it and quit. Else continue with initiate state as the current state.
2. Loop until a solution state is found or until there are no new operators left to be applied in the
current state.
a) Select an operator that has not yet been applied to the current state and apply it to
produce a new state.
b) Evaluate the new state.
i. If it is a goal / solution state, then return it and quit.
ii. If it is not a goal state but it is better than the current state, then make it the current
state
iii. If it is not better than the current state, then continue in the loop.

For example, suppose you are in an unfamiliar city without a map and you want to get downtown. You
simply aim for the tall buildings. The heuristic function is just distance between the current location and
the location of the tall buildings and the desirable states are those in which this distance is minimized.

• Steepest-Ascent Hill Climbing

✓ A variation of simple hill climbing that considers all the moves from the current state and selects
the best one as the next state. This is also called as Gradient search. This contrasts with simple
method in which the first state that is better than current state is taken.

Algorithm Steps:

1. Generate and evaluate the initial state. If it is also a goal state, then return it and quit.
Otherwise, continue with the initial state as the current state.
2. Loop until a solution is found or until a complete iteration produces no change to
current state.
a) Let SUCC be a state such that any possible successor of the current state will be better
than SUCC.
b) for each operator that applies to the current state do:
i) Apply the operator and generate a new state.
ii) Evaluate the new state. If it is a goal state, then return it and quit. If not compare
it to SUCC. If it is better, then set SUCC to this state. If it is not better, leave
SUCC alone.
c) If the SUCC is better than current state, then set current state to SUCC.

✓ Which method will work better for a particular problem?

Depends on: Time required to select a move

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 45

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Number of moves required to set to a solution
✓ Steepest assent takes longer time to select a move
✓ Basic hill climbing takes more number of moves to set to a solution.

Both basic and steepest-Ascent hill climbing may fail to find a solution. They may be terminated by
getting to a state from which no better states can be generated. This can happen, if the program has
reached a local maximum, a plateau, or a ridge.

➔ These are the problems with hill climbing

✓ A local maximum is a state that is better than all its neighbors but is not better than some
other states farther away. At a local maximum, all moves appear to make things worse. Local
maxima are particularly frustrating because they often occur almost within sight of a solution. In
this case, they are called foothills.
✓ A plateau is flat areas of the search space in which the whole set of neighboring states have the
same value on a plateau. It is not possible to determine the direction in which to move.
✓ A ridge is special kind of local maximum. It is an area of the search space that is higher than
surrounding areas, but that cannot be traversed by single moves in any one direction i.e., area has
a slope.

➔ Methods to deal with above problems

✓ Backtrack to some earlier node and try going in a different direction to deal with local maxima
problem. These require maintaining a list of paths almost taken and go back to one of them if the
path that was taken leads to a dead end.
✓ Make a big jump in some direction to try to get to a new section of the search space. This is to
deal with plateaus.
✓ Apply two or more rules before doing the test. This corresponds to moving is several directions
at once. This is a good strategy for dealing with ridges. Even with the above measures, hill
climbing is not always very effective. It is particularly unsuited to problems where the value of
the heuristic function drops off suddenly as you move away from a solution. It is inefficient in a
large and rough problem space.

Example 1: Suppose we may use the following Heuristic

Local: Add one point for every block that is resting on the thing it is supposed to be resting on.
Subtract one point for every block that is sitting on the wrong thing.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 46

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Using this function, the goal state has a score of 8. The initial state has of 4 (since it gets one point
added for blocks C, D, E, F, G and H and one point subtracted for blocks A and B). There is only one
move from the initial state, namely to move block A to the table. That state with a score of 6 (since now
A’s position cause a point to be added rather than subtracted). The hill-climbing procedure will accept
that move. From the new state, there are three possible moves; leading to the three possible moves /
states as shown in Figure above. These states have the scores: (a) 4, (b) 4, and (c) 4. Hill climbing will
halt because all these states have lower scores than the current state. The process has reached a local
maximum that is not the global maximum. The problem is that by purely local examination of support
structures, the current state appears to be better than any of its successors because more blocks rest on the
correct objects. To solve this problem, it is necessary to disassemble a good local structure (the stack B
through H) because it is in the wrong global context.

Example 2: Suppose we may use the following Heuristic

Global: For each block that has the correct support structure (i.e., the complete structure underneath
it is exactly as it should be), add one point for every block in the support structure. For each block that
has an incorrect support structure, subtract one point for every block in the existing support
structure.

Using this function, the goal state has the score 28 (1 for B, 2 for C, …. 7 for H) because all blocks have
correct support structure. The initial state has the score -28 (-1 for C, -2 for D, …. -7 for A, because no
blocks have correct support structure excluding B). Moving A to the table yields a state with a score of
-21 since A no longer have seven wrong blocks under it (-28-(-7)). The three states that can be
produced next now have the following scores: (a) -28, (b) -16 (-15+ (-1) for H), and (c) -15 (H is on
table now). This time, steepest-ascent hill climbing will choose move (c), which is the correct one. This
new heuristic function captures the two key aspects of this problem: incorrect structures are bad and
should be taken apart; and correct structures are good should be build up. As a result, the same hill
climbing procedure that failed with the earlier heuristic function now works perfectly.

• Simulated Annealing
✓ A variation of hill climbing in which, at the beginning of the process, some downhill moves may
be made. Here, an idea is to do enough exploration of the whole space early on, so that, the final
solution is relatively insensitive to the starting state. This may lower the chances of getting caught
at a local maximum, a plateau, or a ridge.
✓ Uses objective function in place of the term heuristic function. Here, attempt is made to
minimize rather than maximize the value of objective function. So it is valley descendent rather
than hill climbing.
✓ Maintains annealing schedule. The rate at which the system is cooled is called the annealing
schedule.
✓ Moves to worse states may be accepted.
✓ Maintains best state so far, so that, if the final state is worse than previous, then earlier state is
still available.
✓ Uses probability formula P = 𝒆−∆𝑬/𝑻 , Where ∆E = change in the value of objective function.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 47

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
✓ Selecting values for T that produces desirable behavior on the part of algorithm.

Simulated annealing as a computational processes patterned after the physical process of annealing, in
which physical substances such as metals are melted (i.e., raised to high energy levels) and then gradually
cooled until some solid state is reached. The goal of this process is to produce a minimal – energy final
state. Thus this process is one of valley descending in which the objective function is the energy level.
Physical substances usually move from higher energy configurations to lower ones, so the valley
descending occurs naturally, but there is some probability that a transition to a higher energy state will
occur. This probability is given by the function P1 = 𝒆−∆𝑬/𝑲𝑻

Where ∆E is the positive change in the energy level, T is the temperature, and K is Boltzmann’s
constant. Thus, in the physical valley descending that occurs during annealing, the probability of a large
uphill move is lower than the probability of a small one. Also, the probability that an uphill move will be
made decreases, as the temperature decreases. One way to characterize this process is that downhill
moves are allowed anytime. Large upward moves may occur early on, but as the process progress, only
relatively small upward moves are allowed until finally the process converges to a local minimum
configuration.

In the physical process, temperature is a well-defined notion, measured in standard units. The variable K
describes the correspondence between the units of temperature and the units of energy. Since, in the
analogous process, the units for both E and T are artificial, it makes sense to incorporate K into T,
selecting values for T that produce desirable behavior on the part of the algorithm. Thus we use the
revised probability formula P = 𝒆−∆𝑬/𝑻 . But we still need to choose a schedule of values for T - a
temperature.

Algorithm Steps:

1. Evaluate the initial state, if it is goal, return it and quit. Else, continue with current state.
2. Initialize BEST-SO-FAR to current state.
3. Initialize T according to annealing schedule.
4. Loop until solution is found, or there are no operators left to be applied in the current state.
a) Choose an operator which is not yet applied, apply it and produce a new state.
b) Evaluate new state, compute
∆E = (value of current) – (value of new state)
i) If new state is goal, return it and quit
ii) If it is not a goal state, but is better than current state, then make new state as current
state
iii) If it is not a better state, make it the current with probability P as defined above.
c) Revise T as necessary according to the annealing schedule.
5. Return BEST-SO-FAR, as the answer.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 48

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
1.14 BEST FIRST- SEARCH
✓ Combines the advantages of both DFS and BFS
✓ Uses behavior to select the most promising paths to the goal node.
✓ At any point in the search process, Best-First moves forward from the most promising of all the
nodes generated so far.
✓ Retains all estimates computed for previously generated nodes and makes its selection based on
the best among all.

• OR- Graphs
✓ At each step of the best-first search process, we select the most promising of the nodes we have
generated so far. This is done by applying an appropriate heuristic function to each of them.
We then expand the chosen node by using the rules to generate its successors. If one of them is
a solution, we can quit. If not, all those new nodes are added to the set of nodes generated so
far. Again the most promising node is selected and the process continues.
✓ Figure below shows the beginning of a best-first search procedure. Initially, there is only one
node, so it will be expanded. Doing so generates three new nodes. The heuristic function,
which, in this example, is an estimate of the cost of getting to a solution from a given node, is
applied to each of these new nodes, since node D is the most promising, it is expanded next,
producing two successor nodes, E and F. But then the heuristic function is applied to them.
Now another path, that going through node B, looks more promising, so it is pursued,
generating nodes G and H. But again when these new nodes are evaluated they look less
promising than another path, so attention is returned to the path through D to E. E is then
expanded, yielding nodes I and J. At the next step, J will be expanded, since it is the most
promising. This process can continue until a solution is found.

✓ Uses two lists of nodes: OPEN and CLOSED.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 49

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
o OPEN – nodes that have been generated and not yet been examined. Heuristic value
computed.
o CLOSED – nodes that have already been examined.
✓ Uses OR graph since each of its branches represents an alternative problem- solving path.

Algorithm Steps: Best - First search

1. Start with OPEN containing just the initial state.

2. Until a goal is found, or there are no nodes left on the OPEN do:
a) Pick the best node on OPEN.
b) Generate its successors
c) For each successor do:
i. If it has not been generated before, evaluate it, add it to OPEN, and record its parent.
ii. If it has been generated before, change the parent if this new path is better than the
previous one. In that case, update the cost of getting to this node and to any
successors that this node may already have.

• Introduction of A* Algorithm
✓ Proposed by Hart in 1972, is a combination of best-first and Branch-and-Bound methods along
with Dynamic Programming principle.
✓ Uses a heuristic or evaluation function f(x) to determine the order in which the search visits
nodes in the tree. For any node N, f(N) is defined as f(N)=g(N) + h(N), where ‘g’ is a measure of
the cost from starting node to N, i.e., sum of costs of the rules that were applied along the best
path to the current node. ‘h’ is an estimate of additional cost of getting from the current node N
to the goal node.
✓ Also called as OR graph / Tree search algorithm.
✓ A* Algorithm incrementally searches all the nodes staring from the start node until it finds the
shortest path to a goal node. Starting with a given node, the algorithm expands the node with the
lowest f(x) value.

A* Algorithm Steps

1. Place the starting node S on OPEN, set f=h, g=0 CLOSED=∅.

2. If OPEN is nil, stop & return failure. Otherwise, continue.
3. Remove a node N from OPEN, which has the least value of f. If node N is a goal node,
return success and stop. Otherwise continue
4. Expand N, generate all successors of N yields N1 and place N on CLOSED.
For each successor N1, if N1 is not already on OPEN, or CLOSED, attach a back pointer
to N. Compute f (N1) and place it on OPEN.
5. If N1 is already on OPEN or CLOSED, should be attached to back pointers which reflect
the lowest of g (N1) path.
If N1 was on CLOSED and its pointer was changed, remove it and place it on OPEN.
6. Return to step 2.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 50

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Consider the situation shown in Figure (a) below. Assume that the cost of all arcs is 1. Initially, all
nodes except A are on OPEN. For each node, f is indicated as the sum of h and g. In this example, node
B has the lowest f = 4, so it is expanded first. Suppose it has only one successor E, which also appears to
be three moves away from a goal. Now f (E) is 5, the same as f(C). Suppose we resolve this in favor of
the path we are currently following. Then we will expand E next. Suppose it too has a single successor
F, also judged to be three moves from a goal. We are clearly using up moves and making no progress.
But f(F) =6, which is greater than f(C). So we will expand C next. Thus we see that by underestimate
h(B), we have wasted some effort. But eventually we discover that B was farther away than we thought
and we go back and try another path.

Now consider the situation shown in Figure (b) below. Again we expand B on the first step. On the
second step we again expand E. At the next step we expand F, and finally we generate G, for a solution
path of length 4. But suppose there is a direct path from D to a solution, giving a path of length 2. We
will never find it. By overestimating h(D), we make D look so bad that we may find some other, worse
solution without ever expanding D.
So, we cannot be guaranteed of finding the cheapest path solution unless we expand the entire graph
until all paths are longer than the best solution.

(a) (b)

• AGENDAS

An agenda is a list of tasks a system could perform. Associated with each task there are usually two
things: a list of reasons why the task is being proposed (often called justifications) and a rating
representing the overall weight of evidence suggesting that the task would be useful.

An agenda-driven system uses the following procedure.

Algorithm: Agenda-Driven Search

1. Do until a goal state is reached or the agenda is empty:

(a) Choose the most promising task from the agenda. This task can be represented in any desired
form. It can be thought of as an explicit statement of what to do next or simply as an
indication of the next node to be expanded.
(b) Execute the task by devoting to it the number of resources determined by its importance.
The important resources to consider are time and space. Executing the task will generate
additional tasks (successor nodes). For each of them, do the following:

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 51

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
(i) See if it is already on the agenda. If so, then see if this same reason for doing it is
already on its list of justifications. If so, ignore this current evidence. If this
justification was not already present, add it to the list. If the task was not on the
agenda, insert it.
(ii) Compute the new task’s rating, combining the evidence from all its justifications. Not
all justifications need have equal weight. It is often useful to associate with each
justification a measure of how strong a reason it is. These measures are then combined
at this step to produce an overall rating for the task.

One important question that arises in agenda-driven systems is how to find the most promising
task on each cycle. One way to do this is- Maintain the agenda sorted by rating. When a new
task is created, insert it into the agenda in its proper place. When a task has its justifications
changed, re-compute its rating and move it to the correct place in the list.

An agenda-driven control structure is also useful if some tasks (or nodes) provide negative
evidence about the merits of other tasks (or nodes). This can be represented by justifications with
negative weightings. If these negative weightings are used, it may be important to check not only
for the possibility of moving a task to the head of the agenda but also of moving a top task to the
bottom if new, negative justifications appear.

Despite of difficulties, Agenda-driven control structures are very useful. They provide an
excellent way of integrating information from a variety of sources into one program since each
source simply adds tasks and justifications to the agenda. As AI programs become more complex
and their knowledge bases grow, this becomes a particularly significant advantage.

1.15 PROBLEM REDUCTION

• AND-OR Graphs

✓ Problem can be decomposed into smaller problems, all of which must then be solved.
✓ AND-OR graph (tree) is useful for representing the solution of problems that are decomposed.
✓ The decomposition generates arcs that we call AND arcs. One AND arc may point to many
number of successor nodes, all of which must be solved in order for the arc point to a solution.
✓ Best – First Search algorithm is not adequate for searching AND-OR graphs.
✓ This uses a value called FUTILITY (threshold cost).

Example:

Figure: A simple AND-OR graph.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 52

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)

• Algorithm: Problem Reduction

1. Initialize the graph G to the starting node.

2. Loop until the starting node is labeled SOLVED or until its cost goes above FUTILITY:
a) Traverse the graph, following the best path, accumulate the set of nodes that are on that
path and have not yet been expanded or labeled as solved.
b) Pick one of these unexpanded nodes and expand it. If there are no successors, assign
FUTILITY as the value of this node. Otherwise, add its successors to the graph and
compute ‘f ’ for each of them. If ‘f ’ of any node = 0, mark it as SOLVED.
c) Change the ‘f’ estimate of the newly expanded node to reflect the new information
provided by its successors. Propagate this change back ward through the graph. If any
node contains a successors are whose descendants are all solved, label the node itself
as SOLVED.

✓ This process is illustrated in Figure below. Assume that the cost of all arcs is 1. At step 1, A is the
only node, so it is at the end of the current best path. It is expanded, yielding nodes B, C and D.
The arc to D is labeled as the most promising one emerging from A, since it costs 6 compared to
B and C, which costs 9. (Marked arcs are indicated in the Figure by arrows). In step 2, node
D is chosen for expansion. This process produces one new arc, the AND arc to E and F, with a
combined cost estimate of 10. So we update the f value of D to 10. Going back one level, we see
that this makes AND arc B-C better than the arc to D, so it is labeled as the current best path.
At step 3, we traverse that arc from A and discover the unexpanded nodes B and C. If we are
going to find a solution along this path, we will have to expand both B and C, and then choose to
explore B first. This generates two new arcs, the ones to G and to H. propagating their f values
backward, we update f of B to 6 (since that is the best we think we can do, which we can achieve
by going through G). This requires updating the cost of the AND arc B-C to 12 (6 + 4 + 2). After
doing that, the arc to D is again the better path from A, so we record that as the current best path
and either node E or node F will be chosen for expansion at step 4. This process continues until
either a solution is found or all paths have led to dead ends, indicating that there is no solution.

✓ The Operation of Problem Reduction

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 53

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)

• AO* Algorithm: For searching AND-OR graphs

✓ Uses a graph, representing the part of the search graph that has been explicitly generated so far.
Each node in the graph will have pointers to immediate successors and procedures.

Algorithm Steps

1. Let graph G consists of only initial node, call it INIT node, and compute h (INIT).
2. Until INIT is labeled SOLVED or until INIT’s ‘h’ value becomes greater than FUTILITY,
repeat the following procedure.
a) Trace marked arcs from INIT & select for expansion, call this as NODE.
b) Generate SUCCESSORS of NODE, if none then assign FUTILITY as ‘h’ value of
NODE and this is not solvable. If SUCCESSORS, then call SUCCESSOR & do the
following.
i. Add SUCCESSOR to G.
ii. If SUCCESSOR is terminated, label as SOLVED and assign ‘h’ to zero.
iii. If SUCCESSOR is not a terminal node, compute ‘h’ value for it.
c) Propagate the newly discovered information up the graph G by doing the following:
Let ‘S’ be the set of of nodes marked SOLVED or whose ‘h’ values have been changed.
Initialize ‘S’ to NODE until S is empty. Repeat the following.
i. Select from ‘S’ a node whose descendent in G occurs in ‘S’. Call this as current
and remove it from ‘S’.
ii. Compute the cost from current, assign as current’s new ‘h’ value minimum.
iii. Mark best path of current by marking arc that had minimum cost.
iv. Mark current SOLVED, if all of nodes connected to it through the new labeled arc
have been labeled SOLVED.
v. If current marked SOLVED, or if cost of current changed, then its new status
propagated. So add all of the ancestors of current to S.

1.16 CONSTRAINT SATISFACTION PROBLEMS (CSP)

✓ A CSP consists of three components: X, D and C

Where: X: is a set of variables {x1, x2, x3 …xn}
D: is a set of domains {D1, D2,….Dn}
C: is a set of constraints that specify allowable combinations of values.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 54

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
✓ Each domain Di consists of a set of allowable values { v1, v2, …….vn} for variable Xi . Each
constraint Ci consists of a pair {scope, rel}, where scope is a tuple of variables that participate in
the constraint and rel is a relation that defines the values that those variables can take on.
✓ To solve CSP, we need to define an assignment of values to some or all of the variables, {xi=vi,
xj=vj…}. An assignment that does not violate any constraints is called a constraint or legal
assignment.
✓ Assignment is Complete if every variable is assigned and solution to CSP is consistent and
Partial if it assigns values to only some of the variables.

• Example 1: Map coloring problem

Map of Australia with its states & territories

GRAPH

NT Q

WA

NSW
SA

✓ The goal is to assign colors to each region so that no neighboring regions have the same color.
✓ To formulate this as CSP, we need to define the variables representing the regions
▪ X= {WA, NT, Q, NSW, V, SA, T}
▪ The domain if the each variable is D𝑖 = {Red, Blue, Green}

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 55

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
✓ The constraints require neighboring regions to have distinct colors. There are nine regions and
hence there are nine constraints.
C= {SA ≠WA, SA≠NT, SA≠Q, SA≠NSW, SA≠V, WA≠NT, NT≠Q, Q≠NSW, NSW≠V}

One possible solution to this problem could be

{WA=Red, NT=Green, Q=Red, NSW=Green, V=Red, SA=Blue, and T=Red}

CSP Domains are: Constraints domains – LPP, searching problems

Discrete domains – map coloring, 8-queens problem
✓ Types of Constraints: Linear, Non-Linear, Binary (relates two Variable), Global (any number of
variables)

• Cryptarithmetic Puzzles
✓ Each letter stands for a distinct digit: 0 to 9 only.
✓ The aim is to find a substitution of digits for letters such that the resulting sum is arithmetically
correct, with the added restriction that no leading zeroes are allowed.

Ex1: T W O
+T W O
FO U R

0+0=R+10.C1
C1+W+W1=U+10.C2
C2+T+T=0+10.C3
C3=F
F=1, → T ≠ 0, and T ≥ 5
T=5 → O=0 O≠0
T=6 → O=2, R=4, W=?
T=7 → O=4, R=8, W=3, U=6 one solution

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 56

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Example: 2

S E N D
+ M O R E
M O N E Y

✓ Constraints
1. Values for letters: 0 to 9 only
2. No two letters should have the same value
3. Sum of the digits much be as shown in the problem.

✓ Initial state

S=? M=? C1=?

E=? 0=? C2=?
N=? R=? C3=?
D=? E=? C4=?

✓ Carry: C4 C3 C2 C1

Solution

1. M=1, since two single digit numbers plus a carry cannot total more than 19.
2. S=8 or 9, since S+M+C3>9 and M=1, S+1+C3>9, so S+C3>8 and c3 is at most1.
3. O= ∅ , since S+M (1) +C3 (≤1) must be at least 10 to generate a carry and it can be at most 11.
But M is already 1, so O must be 0(zero).
4. N=E or E+1, depending on the C2 value. But N cannot have the same value as E, so, N=E+1 and
C2=1.
5. In order for C2 to be 1, the sum of N+R+C1 must be greater than 9. So, N+R must be greater than
8.
6. N+R cannot be greater than 18; even with a carry in, so, E cannot be 9.
At this point, let us assume that no more constraints can be generated. Then to make progress
from here, we must guess. Suppose E=2, (this occurs three times) N=3 (E+1) and R= 8 or 9 since
R +N(3)+C1 (1 or 0)= 2 or 12. But N is already 3, thus R+3+ (0 or 1) =12 and R=8 or 9.
→ 2+D=y or 2+D=10+y, from the sum in the right most column. Again assuming no further
constraints can be generated a guess is required for C1. Suppose that C1=1, then R=9 and 2+D=Y
Thus 8 or 9 is assigned to R or D or S

✓ One possible solution would be:

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 57

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)

SEND with E=2, S=9 S=8

MORE 9235 823?
MONEY 1082 1092
10 3 2 1032

E=4 E=5
E=3 945 ? 9567
9347 1084 1 085
1083 1054 10652
1043

C3=0, C1=1, C2=1

S=9, E=5, N=6, D=7, M=1, O = ∅, R=8 and Y=2

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 58

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Similarly try for

3. BASE 4. R O A D S 5. D O N A L D
+BALL + CROSS +GERALD
GAMES DANGER ROBERT

6. L O G I C 7. D R I N K ANS for 4 QUE: 62513

+LOGIC +DANCE 96233
P ROLOG PARTY 158746

Constraints satisfaction is a search procedure that operates in a space of Constraint sets. The initial state
contains the constraints that are originally given in the problem description. A Goal State is any state
that has been constrained “enough”, where “enough” must be defined for each problem. For example,
for crypt arithmetic, enough means that each letter has been assigned a unique numeric value.

Constraint satisfaction is a two-step process. First, Constraints are discovered and propagated as far as
possible throughout the system. Then, if there is still not a solution, search begins. A guess about
something is made and added as a new Constraint. Propagation can then occur with this new Constraint,
and so forth.

Algorithm: Constraint Satisfaction

1. Propagate available constraints. To do this, first set OPEN to the set of all objects that must have
values assigned to them in a complete solution. Then do until an inconsistency is detected or until
OPEN is empty:
a) Select an object OB from OPEN. Strengthen as much as possible the set of constraints that
apply to OB.
b) If this set is different from the set that was assigned the last time OB was examined or if this
is the first time OB has been examined, then add to OPEN all objects that share any
constraint with OB.
c) Remove OB from OPEN.
2. If the union of the constraints discovered above defines a solution, then quit and report the
solution.
3. If the union of the constraints discovered above defines contradiction, then return failure.
4. If neither of the above occurs, then it is necessary to make a guess at something in order to
proceed. To do this, loop until a solution is found or all possible solutions have been eliminated:
a) Select an object whose value is not yet determined and select a way of strengthening the
constraints on that object.
b) Recursively invoke constraint satisfaction with the current set of constraints augmented by the
strengthening constraint just selected.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 59

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
This algorithm has been stated as generally as possible. To apply it in a particular problem
domain requires the use of two kinds of rules: rules that define the way constraints may
validly be propagated and rules that suggest guesses when guesses are necessary.

1.17 MEANS-ENDS-ANALYSIS

✓ This process centers on the detection of differences between the current state and the goal
state. Once the difference is isolated, an operator that can reduce the difference can be found.
✓ The problem space of means-ends-Analysis has an initial state (object) and one or more goal
states (objects), a set of operators Ok with given pre-conditions for their applications and a
difference function that computes the difference between two states Si and Sj

Algorithm steps

✓ Compare the current state ‘Si’ and goal state ‘Sj’ compute the difference Dij.
1. Select an operator ok to reduce the difference Dij.
2. Apply the ok if possible. Otherwise, save the current state, create the subgoal and apply
means-ends-Analysis recursively to reduce the subgoal.
3. If the subgoal is solved, the saved state is restored and work is resumed on the original
problem.

Example 1: Consider a simple household robot domain. The available operators are shown in Figure
(a) below, along with their preconditions and results. Figure (b) shows the difference table that describes
when each of the operators is appropriate. Notice that sometimes there may be more than one operator
that can reduce a given difference and that a given operator may be able to reduce more than one
difference.

Suppose that the robot in this domain were given the problem of moving a desk with two things on it
from one room to another. The objects on top must also be moved. The main difference between the start
state and the goal state would be the location of the desk. To reduce this difference, either PUSH or
CARRY could be chosen. If CARRY is chosen first, its preconditions must be met. These results in two
more differences that must be reduced: the location of the robot and the size of the desk. The location
of the robot can be handled by applying WALK, but there are no operators that can change the size of an
object. So this path leads to a dead-end. Following the other branch, we attempt to apply PUSH. Figure
(c) shows the problem solver’s progress at this point. It has found a way of doing something useful. But it
is not yet in a position to do that thing. And the thing does not get it quite to the goal state. So now the
differences between A and B and between C and D must be reduced.

PUSH has four preconditions, two of which produce differences between the start and the goal states: the
robot must be at the desk, and the desk must be clear. Since the desk is already large, and the robot’s arm
is empty, those two preconditions can be ignored. The robot can be brought to the correct location by
using WALK. And the surface of the desk can be cleared by two uses of PICKUP. But after one
PICKUP, an attempt to do the second results in another difference – the arm must be empty.
PUTDOWN can be used to reduce that difference.

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 60

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Once PUSH is performed, the problem state is close to the goal state, but not quite. The objects must be
placed back on the desk. PLACE will put them there. But it cannot be applied immediately. Another
difference must be eliminated, since the robot must be holding the objects. The progress of the problem
solver at this point is shown in Figure (d) below.

Figure (a): The Robot’s Operators

Operator Preconditions Results

at (robot, obj) ^
large (obj) ^ at (obj, loc) ^
PUSH (obj, loc)
Clear (obj) ^ at (robot, loc)
Armempty
At (robot, obj)^ At (obj, loc)^
CARRY(obj, loc)
Small (obj) At (robot, loc)
WALK(loc) None At (robot, loc)
PICKUP(obj) At (robot, obj) Holding (obj)
PUTDOWN(obj) Holding (obj) Holding (obj)
At (robot, obj2) ^
PLACE(obj1, obj2) On (obj1,obj2)
Holding (obj1)
Figure (b): A Difference Table

Figure (c): Progress of the Means-Ends Analysis Method

Figure (d): More Progress of the Means-Ends Analysis Method

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 61

 ARTIFICIAL INTELLIGENCE & MACHINE LEARNING (18CS71)
Example 2: Given an initial sate and the goal sate the system attempts to transform the initial state to the
goal state through series of operator application transformations.

Example:

Initial state: R^ (~ p→ 𝑞)

Final state: (q ⋁ p) ^ R

Operators: 1. A ⋁ B ⟺ B ⋁ 𝐴

2. A→ 𝐵 ⟺ ~ A⋁B
3. A ⋀ B ⟺ 𝐵⋀𝐴
4. (A→ 𝐵) ⟺ ~ B → ~ A

Difference:

R ⋀ (~ p→ 𝑞)
(q ⋁ p) ^ R

Initial state: R^ (~ p→ 𝑞)
Apply A ⋀ B ⟺ 𝐵⋀ 𝐴
⟹ (~ p→ 𝑞) ^ R
Apply A→ 𝐵 ⟺ ~ A⋁B
⟹ (p⋁q) ^ R
Apply A⋁B ⟺ B ⋁ 𝐴
⟹ (q ⋁ p) ^ R goal state.

******************************************************

SDV, SGN,PR, Dept. of ISE, RNSIT 2022-2023 Page 62