Machine Learning

(CSO851)

Lecture - 01
Acknowledgement

 https://www.cmpe.boun.edu.tr/~ethem/i2ml3e/
 https://www.cs.ox.ac.uk/people/nando.defreitas/machinelearning/
 https://nptel.ac.in/courses/106/106/106106202/
Overview of the Course
 Introduction to Machine Learning
 Foundations of Machine Learning
 Machine Learning Algorithms
 Regressions (Linear, Polynomial, Step-wise)
 Logistic Regression
 Gradient Descent
 Decision Trees
 Bayes Theorem and Bayes Classification
 Support Vector Machines
 Principal Component Analysis
 Linear Discriminant Analysis
 Nearest Neighbors
 Density Estimation
 Neural Networks
Evolution of Machine Learning
 1950 — Alan Turing creates the “Turing Test” to determine if
a computer has real intelligence. To pass the test, a computer
must be able to fool a human into believing it is also human.
 1952 — Arthur Samuel wrote the first computer learning
program. The program was the game of checkers.
 1957 — Frank Rosenblatt designed the first neural network for
computers (the perceptron), which simulate the thought
processes of the human brain.
 1967 — The “nearest neighbor” algorithm was written,
allowing computers to begin using very basic pattern
recognition.
 1979 — Students at Stanford University invent the “Stanford
Cart” which can navigate obstacles in a room on its own.
https://www.forbes.com/sites/bernardmarr/2016/02/19/a-short-history-of-machine-learning-every-
manager-should-read/?sh=413d1dcc15e7
Evolution of Machine Learning
 1981 — Gerald Dejong introduces the concept of Explanation Based Learning
(EBL), in which a computer analyses training data and creates a general rule
it can follow by discarding unimportant data.
 1985 — Terry Sejnowski invents NetTalk, which learns to pronounce words
the same way a baby does.
 1990s — Work on machine learning shifts from a knowledge-driven approach
to a data-driven approach.
 1997 — IBM’s Deep Blue beats the world champion at chess.
 2006 — Geoffrey Hinton coins the term “deep learning” to explain new
algorithms that let computers “see” and distinguish objects and text in
images and videos.
 2010 — The Microsoft Kinect can track 20 human features at a rate of 30
times per second, allowing people to interact with the computer via
movements and gestures.
Evolution of Machine Learning
 2011 — IBM’s Watson beats its human competitors at Jeopardy.
 2011 — Google Brain is developed, and its deep neural network can learn to
discover and categorize objects much the way a cat does.
 2012 – Google’s X Lab develops a machine learning algorithm that is able to
autonomously browse YouTube videos to identify the videos that contain
cats.
 2014 – Facebook develops DeepFace, a software algorithm that is able to
recognize or verify individuals on photos to the same level as humans can.
 2015 – Amazon launches its own machine learning platform.
 2015 – Microsoft creates the Distributed Machine Learning Toolkit.
 2016 – Google’s artificial intelligence algorithm beats a professional player
at the Chinese board game Go, which is considered the world’s most
complex board game and is many times harder than chess.
Introduction to Machine Learning
 Machine Learning: According to Arthur Samuel, Machine Learning algorithms
enable the computers to learn from data, and even improve themselves, without
being explicitly programmed.
 A computer program is said to learn from experience E with respect to some class of tasks
T and performance measure P, if its performance at tasks in T, as measured by P, improves
with experience E.
 A checkers learning problem:
 Task T: playing checkers
 Performance measure P: percent of games won against opponents
 Training experience E: playing practice games against itself
 A handwriting recognition learning problem:
 Task T: recognizing and classifying handwritten words within images
 Performance measure P: percent of words correctly classified
 Training experience E: a database of handwritten words with given classifications
When Learning is Needed?

 Human expertise does not exist (navigating on Mars).

 Humans are unable to explain their expertise (speech recognition).
 Solution changes in time (stock market, routing on a computer network).
 Solution needs to be adapted to particular cases (user biometrics)

Many tasks cannot be adequately solved using a simple

(deterministic), rule-based solution. A large number of
factors could influence the answer.
Applications
 Healthcare – Machine learning solutions are being used to help medical professionals detect
the presence of diseases, and machine learning processes are also used to detect emotional
states via smartphone data.
 Retail – Machine learning technology powers Amazon’s “You May Also Like” suggestions, and
machine learning processes are also behind Netflix’s recommendation algorithm.
 Manufacturing – Machine learning helps manufacturers reduce process-driven losses,
increase capacity by optimizing the production process, and reduce costs by guiding
predictive maintenance.
 Banking & Finance – Banks and financial institutions utilize machine learning to detect and
prevent fraudulent transactions, and to identify insights in financial data, allowing them to
find better investment opportunities, advise investors about when to trade, and provide
other important insights related to financial matters.
 Transportation – Both Google and Tesla’s self-driving cars are powered by machine learning
technology, including deep learning which allows them to interpret, predict and respond to
data critical for autonomous driving.
https://csuglobal.edu/blog/why-is-machine-learning-so-important
Types of Machine Learning
 Association Learning: Association learning is a rule based machine learning and data
mining technique that finds important relations between variables or features in a
data set.
 Supervised Learning: Supervised learning is one of the most basic types of machine
learning. In this type, the machine learning algorithm is trained on labeled data.
Supervised learning is extremely powerful when used in the right circumstances.
 Unsupervised Learning: There are no labels or correct outputs. The task is to discover
the structure of the data: for example, grouping similar items to form “clusters”, or
reducing the data to a small number of important “dimensions”.
 Semi-supervised Learning: Semi-supervised learning is an approach to machine
learning that combines a small amount of labeled data with a large amount of
unlabeled data during training. Examples - GAN
 Reinforcement Learning: Reinforcement learning directly takes inspiration from how
human beings learn from data in their lives. It features an algorithm that improves
upon itself and learns from new situations using a trial-and-error method. Favorable
outputs are encouraged or ‘reinforced’, and non-favorable outputs are discouraged or
‘punished’.
https://www.potentiaco.com/what-is-machine-learning-definition-types-applications-and-examples/
Practical Uses of Association Learning
 Basket data analysis – Whether planning product placement in a storefront, running
a marketing campaign or designing a business catalog, association mining is a useful
tool to take the guesswork out of what your customers are looking for.
 Web usage mining and intrusion detection – Finding these hidden correlations is a
powerful predictive tool to discover brand new security threats and network
performance issues that haven’t been analyzed first by a human.
 Bioinformatics – Pre-processing of protein interaction networks for the task of
protein function prediction. To reveal the relationships between miRNAs and
different diseases/phenotypes.

P (Y|X) probability that somebody who buys X also buys Y where X and Y are
products/services.
Example: P (coffee|snacks) = 0.7
Moderna was one of the first to introduce a COVID-19 vaccination during COVID
period. Moderna utilized artificial intelligence to aid in the development of
mRNA sequences.
Practical Uses of Supervised Learning: Classification
 Differentiating between low-risk and high-
risk customers from their income and savings

Discriminant: IF income > θ1 AND savings > θ2

THEN low-risk ELSE high-risk
Practical Uses of Supervised Learning: Classification
 Pattern recognition
 Face recognition: Pose, lighting, occlusion (glasses, beard), make-up, hair
style
 Character recognition: Different handwriting styles.
 Speech recognition: Temporal dependency.
 Medical diagnosis: From symptoms to illnesses
 Biometrics: Recognition/authentication using physical and/or behavioural
characteristics: Face, iris, signature, etc.
Practical Uses of Supervised Learning: Regression
Regression analysis is a
set of statistical
methods used to
estimate relationships
between a dependent
variable and one or
more independent
variables.

https://skepticalscience.com/its-a-climate-shift-step-function-caused-by-natural-cycles.htm
Practical Uses of Unsupervised Learning
Practical Uses of Reinforcement Learning
 A great example is the use of AI agents by Deepmind
to keep cool Google Data Centers. This led to a 40%
reduction in energy spending. The centers are fully
controlled with the AI system without the need for
human intervention. There is obviously still
supervision from data center experts. The system
works in the following way:

 Taking snapshots of data from the data centers every

five minutes and feeding this to deep neural networks
 It then predicts how different combinations will affect
future energy consumptions
 Identifying actions that will lead to minimal power
consumption while maintaining a set standard of
safety criteria
 Sending and implement these actions at the data
center

https://deepmind.com/blog/article/safety-first-ai-autonomous-data-centre-cooling-and-industrial-control
Practical Uses of Reinforcement Learning
 Industry automation
 Trading and finance
 Natural Language Processing
 Healthcare
 Engineering
 News recommendation
 Gaming
 Marketing and advertising
 Robotics manipulation

https://neptune.ai/blog/reinforcement-learning-applications
Foundations of Machine Learning
 The problem of learning
 Basic Terminologies
 Unknown functions
 Training and Selection of Models
 Design a learning system
 Determine the training experience
 Determine the target function
 Determine a representation for target function
 Determine a function approximation algorithm
 The final design
The Problem of Learning & Basic Terminology
 There are a known set and an unknown function on . Given data, construct a
good approximation of . This is called learning .
 Domain: The set called feature space and an element is called feature
vector or an input. The coordinates of are called features. Individual features
may take values in a continuum, a discrete, ordered set or a discrete, unordered
set.
 Range: The range is usually either a finite, unordered set, in which case
learning is called classification or it is a continuum, in which case learning is called
regression. An element is called a class in classification and a response in
regression.
 Data: In principle, data are random draws from a probability distribution
on . Depending on the problem at hand, the data which are observed may
either consist of domain-range pairs or just a domain values : learning is
called supervised in the former case and unsupervised in the later case.
Basic Terminology
 In supervised learning, the data are

 where each is drawn from a joint probability distribution P(X, Y) on .

Such data are called marked data.
 In unsupervised learning, the data are

Such data are called unmarked data. The range is assumed to be finite, in which
case unsupervised learning is called clustering.
 If the range is [0, ) and the function f to be learned is the mass or density
function of the marginal distribution of the features,

in which case unsupervised learning is called density estimation.

Classification of Learning Problems

Four categories of learning problems

Basic Terminology
 Loss: Good approximation is specified by a loss function

where for each (x, y) ,, is the loss, or penalty, incurred by

approximating y with . Common examples are