MATH -260 Course @ Sorbonne

Day 1: Multivariate Data Analysis∗

Dr. Tanujit Chakraborty, Ph.D. from ISI Kolkata.

Assistant Professor in Statistics at Sorbonne University.
[email protected]
https://www.ctanujit.org/MDA.html
Course for BSc Mathematics (Spe. Data Science) Students.
————————————————————————————-
∗
In 1962, John Tukey described a field called "data analysis" which resembles modern data science.
 Quote of the day
“It is very easy to be a teacher, but very difficult to be a student.

A good student has to learn many concepts, perform in examinations, loyal to his /
her teacher and others.”

2/92
 Why this course?

3/92
 Course Prerequisites

Probability
Programming Theory:
Basic Statistics: Skills:
• Introductory
• Descriptive Statistics • Basics of R & RStudio Probability Theory

• Estimation and • Basic statistics using R • Probability

Hypothesis Testing Distributions

• Analysis of Variance • Basics of Python • Sampling

(ANOVA) Distributions
• Basic statistics using
Python • Linear Algebra &
Calculus

4/92
 Course Syllabus

Regression Multivariate Coding & Projects:

Analysis: Analysis: • Implementations in
• Linear Regression • SVD & PCA RStudio (mostly) &
Python
• Nonlinear Regression • Factor Analysis
• Kaggle Competition
• Shrinkage Methods • LDA & QDA (20%)

• Logistic Regression & • Correspondence • Project Presentation

GLM Analysis (20%)

• Time Series • Market Basket • Mid-term (20% and

Forecasting Analysis Final Exam (40%)

5/92
 What is Data?

“Data is the new oil. It’s valuable, but if unrefined it cannot really
be used. It has to be changed into gas, plastic, chemicals, etc to
create a valuable entity that drives profitable activity; so must data
be broken down, analyzed for it to have value.”

- Clive Humby, UK Mathematician and Architect of Tesco’s Clubcard.

Example:

6/92
 Role of data: Present

7/92
 Role of data: Present

“It is a capital mistake to theorize before one has data. In-

sensibly one begins to twist facts to suit theories, instead of
theories to suit facts.”

- Sherlock Holmes, in A Scandal in Bohemia.

8/92
 The World is Data Rich

9/92
 Recent Trends and Buzzwords

1 Statistics is the study of the collection, analysis, interpretation,

presentation and organization of data.

2 Data science is the study of the generalizable extraction of knowledge

from data, yet the key word is science.

3 Machine learning is the sub-field of computer science that gives

computers the ability to learn without being explicitly programmed.

4 Artificial Intelligence research is defined as the study of intelligent

agents: any device that perceives its environment and takes actions that
maximize its chance of success at some goal.

5 Big data is an all-encompassing term for any collection of data sets so

large and complex that it becomes difficult to process using traditional
data processing applications.

10/92
 T OPIC 1 : S TATISTICS

11/92
 Statistics: View Points

“Statistics is the universal tool of inductive inference, research in natural and so-
cial sciences, and technological applications. Statistics must have a clearly defined
purpose, one aspect of which is scientific advance and the other, human welfare and
national development”
- Professor P C Mahalanobis.
“All knowledge is, in final analysis, History.
All sciences are, in the abstract, Mathematics.
All judgements are, in their rationale, Statistics.”
- Professor C R Rao.

• Role of Statistics:
1 Making inference from samples
2 Development of new methods for complex data sets
3 Quantification of uncertainty and variability
• Two Views of Statistics:
1 Statistics as a Mathematical Science
2 Statistics as a Data Science

12/92
 Statistics: View Points

In 1925, R A Fisher published Statistical Meth-

ods for Research Workers in the Biological Mono-
graphs and Manuals Series by the publisher Oliver
and Boyd of Edinburgh in Scotland. In the first
few pages of the book Fisher gives an Introduction
to statistics and its methods. We give below a part
of this Introduction:-
The science of Statistics is essentially a branch
of Applied Mathematics, and may be regarded as
mathematics applied to observational data. As
in other mathematical studies, the same formula
is equally relevant to widely different groups of
subject-matter. Consequently the unity of the dif-
ferent applications has usually been overlooked, the
more naturally because the development of the un-
derlying mathematical theory has been much ne-
glected. Statistics may be regarded as (i) the study
of populations, (ii) as the study of variation, (iii) as
the study of methods of the reduction of data.
13/92
 History of Statistics : Pre 1900

• Probability and its application to Gambling &

Astronomy

- Jacob Bernoulli (Prob. distribution, Law of

large numbers, etc.)
- Pierre-Simon Laplace (Double exponential,
Transformation, etc.)
- Thomas Bayes (Bayes’ theorem, etc.)
- Gauss and Legendre (Least Square Method)
- Francis Galton (Correlation & Regression)
- Karl Pearson (χ2 test, distribution, etc.)
- and many others.

14/92
 History of Statistics : 1900 - 1980

• Development of Statistics and its application to

Agriculture, Economics, Geology, Medical,
Technology, Clinical Trials, etc.

- Ronald Fisher (Discriminant analysis,

Likelihood, ANOVA & DOE, etc.)
- Jerzy Neyman, Egon Sharpe Pearson & Wald
(Decision theory, Optimality, etc.)
- Lehmann, Hotelling, Anderson & Tukey
(Multivariate & Inferential Statistics, etc.)
- Box, Cox, Jenkin and Blackwell (Time Series)
- Shewhart, Deming, Taguchi & Juran (SQC)
- Efron, Breiman, Friedman, Cramer (Modern
Statistical Tools)
- PC Mahalanobis (Mahalanobis Distance), C. R.
Rao (Linear Models, Multivariate Analysis,
Orthogonal arrays, etc.), etc.
15/92
 History of Statistics : Upto 1980

• Parametric Models : One Sample, two sample,

linear models, survival data, Estimation, Testing
of Hypothesis.
• Probability distributions were believed to
generate data (e.g., Gaussian, Logistic, Poisson,
Exponential, etc.).
• Semiparametric & Nonparametric Models :
Dropping assumptions on population,
dependence and errors.
• Emphases on Optimality in various ways :
Bayes optimality, Decision theory, minimax and
unbiasedness.
• Exact distributional (t, F) approaches and
asymptotic methods (samples size → ∞ viewed
as approximation).

16/92
 Statistics : 1980 - Present

• Data : Large bodies of data with complex data structures are generated
from computers, sensors, manufacturing industries, etc.

• Models : Non/Semiparametric models but in complex probability

spaces / high-dimensional functional spaces (e.g., deep neural net,
reinforcement learning, decision trees, etc.).

• Emphases : Making predictions, causation, algorithmic convergence.

• Data are necessary and at the core of Statistical Learning, Data Science
& Machine Learning.

• Statistics : Not only has strong interactions with Probability but also
other parts of Data Science (Machine Learning, Artificial Intelligence,
etc.).

17/92
 Statistics : Present (2021)

• Probability : Has moved to the center of Mathematics and having strong

interactions with Statistical Physics and Theoretical Computer Science.

• Statistics : Not only has strong interactions with Probability but also
other parts of Data Science (Machine Learning, Artificial Intelligence,
etc.).

• Computational : Computing skills are essential, construction of fast

training algorithms and computation time.

• Applications : Strong interactions with substantive fields in all areas.

Applications of statistical methods in almost all the fields are evident.
Statistics became a key technology driven by data (“Data is the new
oil").

18/92
 Statistics: Past Vs. Present

• Traditional Problems in Applied Statistics:

- Well formulated question that we would like to answer.
- Expensive to gather data and/or expensive to do computation.
- Create specially designed experiments to collect high quality data.

• Current Situation : Information Revolution

- Improvements in computers and data storage devices.
- Powerful data capturing devices.
- Lots of data with potentially valuable information available.

19/92
 What is the Difference?

• Data characteristics:
- Size
- Dimensionality
- Complexity
- Messy
- Secondary sources

• Focus on generalization performance :

- Prediction on new data
- Action in new circumstances
- Complex models needed for good generalization

• Computational considerations :
- Large scale and complex systems

20/92
 T OPIC 2 : D ATA S CIENCE

21/92
 What is Data Science?

“Data is the new oil. It’s valuable, but if unrefined it cannot

really be used. It has to be changed into gas, plastic, chemicals,
etc to create a valuable entity that drives profitable activity; so
must data be broken down, analyzed for it to have value.”

- Clive Humby, UK Mathematician and Architect of Tesco’s

Clubcard.

“Everybody needs data literacy, because data is everywhere.

It’s the new currency, it’s the language of the business. We need
to be able to speak that.”
- MIT Sloan School of Management

22/92
 What is Data Science?

• Interdisciplinary field about scientific methods, processes, and

systems to extract knowledge or insights from complex data in
various forms, either structured or unstructured.
• A concept to unify Statistics, Data Analysis and their related
methods in order to “understand and analyze actual phenomena”
with data.
• Employs techniques and theories drawn from many fields within
the broad areas of Mathematics, Statistics, Information Science,
and Computer Science, in particular from the subdomains of
Machine learning, classification, cluster analysis, data mining,
databases, and visualization.
• Fourth paradigm of Science (empirical, theoretical, computational
and data-driven)

23/92
 Types of Data Science?
"When you’re fundraising, it’s AI.
When you’re hiring, it’s ML.
When you’re implementing, it’s Linear Regression.
When you’re debugging, it’s printf()."
- Baron Schwartz, Founder and CEO of VividCortex, 2017.

24/92
 Data in Data Analytics

Basic Definitions:
Entity: A particular thing is called entity or object.
Attribute: An attribute is a measurable or observable property of an entity.
Data: A measurement of an attribute is called data.
Note: Data defines an entity and Computer can manage all type of data (e.g.,
audio, video, text, etc.). In general, there are many types of data that can be used
to measure the properties of an entity.
Scale: A good understanding of data scales (also called scales of measurement) is
important. Depending on the scales of measurement, different techniques are fol-
lowed to derive hitherto unknown knowledge in the form of patterns, associations,
anomalies or similarities from a volume of data.

25/92
 NOIR: Scales of Measurement
• The NOIR scale is the fundamental building block on which the extended data
types are built.
• Further, nominal (Blood groups, Attendance) and ordinal (Shirt size) are
collectively referred to as categorical or qualitative data. Whereas, interval
(weight, temperature) and ratio (Sound intensity in Decibel) data are collectively
referred to as quantitative or numeric data.

26/92
 Data Cube

Concept of data cube:

A multidimensional data model views data in the form of a cube. A data cube is
characterized with two things:

• Dimension: The perspective or entities with respect to which an organization

wants to keep record.
• Fact: The actual values in the record.

Example: Rainfall data of Meteorological Department.

• Time (Year, Season, Month, Week, Day, etc.)

• Location (Country, Region, State, etc.)

27/92
 2-D view of rainfall data

• In this 2-D representation, the rainfall for “North-East” region are shown with
respect to different months for a period of years...

28/92
 View of 3-D rainfall data

• Suppose, we want to represent data according to times (Year, Month) as well as

regions of a country say East, West, North, North-East, etc.

Figure: 2-D view of rainfall data Figure: 3-D view of rainfall data

29/92
 Data Mining Process

30/92
 Data Mining Process

31/92
 1. Prior Knowledge

Gaining information on
• Objective of the problem.
• Subject area of the problem.
• Data.

32/92
 2. Data Preparation

Gaining information on
• Data Exploration and Data quality.
• Handling missing values and Outliers.
• Data type conversion.
• Transformation, Feature selection and Sampling.

33/92
 3. Modeling

Figure: Splitting data into training and test data sets (right).

34/92
 3. Modeling

Figure: Evaluation of test dataset (right).

35/92
 Application and Knowledge

4. Application:
• Product readiness.
• Technical integration.
• Model response time.
• Remodeling.
• Assimilation.

5. Knowledge:
• Posterior knowledge.

Objectives of Data Exploration:

• Understanding data.
• Data preparation and Data mining tasks.
• Interpreting data mining results.

36/92
 Data exploration

Roadmap:
• Organize the data set.
• Find the central point for each attribute (central tendency).
• Understand the spread of the attributes (dispersion).
• Visualize the distribution of each attributes (shapes).
• Pivot the data.
• Watch out for outliers.
• Understanding the relationship between attributes.
• Visualize the relationship between attributes.
• Visualization high dimensional data sets.
• For more details, read Kotu, V., & Deshpande, B. (2014). Predictive analytics and
data mining: concepts and practice with rapidminer. Morgan Kaufmann.

37/92
 Overview of Data Science Tools

38/92
 T OPIC 3 : M ACHINE L EARNING

39/92
 What is Machine Learning?
Machine learning is the field of study that gives computers the ability to
learn without being explicitly programmed.

40/92
 ML techniques are impacting our life

A day in our life with ML techniques...

41/92
 Introduction to Machine Learning

• Designing algorithms that ingest data and learn a model of the data.
• The learned model can be used to

1 Detect patterns/structures/themes/trends etc. in the data

2 Make predictions about future data and make decisions

• Modern ML algorithms are heavily "data-driven".

• Optimize a performance criterion using example data or past
experience.
42/92
 Types of Machine Learning

• Unsupervised Learning:
- Uncover structure hidden in ‘unlabelled’ data.
- Given network of social interactions, find communities.
- Given shopping habits for people using loyalty cards: find groups of
‘similar’ shoppers.
- Given expression measurements of 1000s of genes for 1000s of
patients, find groups of functionally similar genes.
- Goal: Hypothesis generation, visualization.

• Supervised Learning:
- A database of examples along with ‘labels’ (task-specific).
- Given expression measurements of 1000s of genes for 1000s of patients
along with an indicator of absence or presence of a specific cancer,
predict if the cancer is present for a new patient.
- Given network of social interactions along with their browsing habits,
predict what news might users find interesting.
- Goal: Prediction on new examples.

43/92
 Types of Machine Learning

• Semi-supervised Learning:
- A database of examples, only a small subset of which are labelled.

• Multi-task Learning:
- A database of examples, each of which has multiple labels
corresponding to different prediction tasks.

• Reinforcement Learning:
- An agent acting in an environment, given rewards for performing
appropriate actions, learns to maximize its reward.

44/92
 A Typical Supervised Learning Workflow

Supervised Learning: Predicting patterns in the data

45/92
 A Typical Unsupervised Learning Workflow

Unsupervised Learning: Discovering patterns in the data

46/92
 A Typical Reinforcement Learning Workflow
Reinforcement Learning: Learning a ”policy" by performing
actions and getting rewards (e.g, robot controls, beating games)

47/92
 Machine Learning: A Brief Timeline

48/92
 Machine Learning in the real-world
Broadly applicable in many domains (e.g., internet, robotics, healthcare and
biology, computer vision, NLP, databases, computer systems, finance, etc.).

49/92
 Machine Learning helps NLP

ML algorithm can learn to translate text in the domain of

Natural Language Processing

50/92
 Machine Learning helps Computer Vision

• Automatic generation of text captions for images:

A convolutional neural network is trained to interpret images, and its output is
then used by a recurrent neural network trained to generate a text caption.
• The sequence at the bottom shows the word-by-word focus of the network on
different parts of input image while it generates the caption word-by-word.

51/92
 ML helps Recommendation systems
• A recommendation system is a machine-learning system that is based on data
that indicate links between a set of a users (e.g., people) and a set of items (e.g.,
products).
• A link between a user and a product means that the user has indicated an interest
in the product in some fashion (perhaps by purchasing that item in the past).
• The machine-learning problem is to suggest other items to a given user that he
or she may also be interested in, based on the data across all users.

52/92
 Machine Learning helps Chemistry
ML algorithms can understand properties of molecules and learn to synthesize new
molecules.

Figure: Inverse molecular design using machine learning: Generative models for
matter engineering (Science, 2018)
53/92
 Machine Learning helps Image Recognition

54/92
 ML helps Many Other Areas...

55/92
 Taxonomy of Machine Learning

56/92
 Cheatsheet

57/92
 Development of Stat & ML Models
• Random forest (Breiman,
• Linear Regression 2001).
• ARIMA Model (Box
(Galton, 1875).
and Jenkins, 1970). • Deep Convolutional
• Linear Neural Nets (Krizhevsky,
• Classification and
Discriminant Sutskever, Hinton, NIPS
Regression Tree
Analysis (R.A. 2012).
(Breiman et al., 1984).
Fisher, 1936).
• Generative Adversarial
• Artificial Neural
• Logistic Regression Nets (Ian Goodfellow et
Network (Rumelhart
(Berkson, JASA, al., NIPS 2014).
et al., 1985).
1944).
• Deep Learning (LeCun,
• MARS (Friedman,
• k-Nearest Neighbor Bengio, Hinton, Nature
1991, Annals of
(Fix and Hodges, 2015).
Statistics).
1951).
• Bayesian Deep Neural
• SVM (Cortes and
• Parzen’s Density Network (Yarin Gal,
Vapnik, Machine
Estimation (E Islam, Zoubin
learning, 1995)
Parzen, AMS, 1962) Ghahramani, ICML
2017).

58/92
 Developments of Neural Nets

Fig: Developments of Neural Network Models

59/92
 Supervised Learning: Classification
• Example: Credit scoring.

• 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.

• Classification: Learn a linear/nonlinear

separator (the “model") using training
data consisting of input-output pairs
(each output is discrete-valued “label" of
the corresponding input).

• Use it to predict the labels for new “test"

inputs.

• Other Applications: Image Recognition,

Spam Detection, Medical Diagnosis.

60/92
 Supervised Learning: Regression

• Example: Price of a used car.

• X : car attributes; Y : price and

Y = f(X, θ)

• f( ) is the model and θ is the model

parameters.

• Regression: Learn a line/curve (the

“model") using training data consisting
of Input-output pairs (each output is a
real-valued number).

• Use it to predict the outputs for new

“test" inputs.

• Other Applications: Price Estimation,

Process Improvement, Weather
Forecasting.

61/92
 Unsupervised Learning: Clustering

• Given: Training data in form of

unlabeled instances
{x1 , x2 , ..., xN }

• Goal: Learn the intrinsic latent

structure that
summarizes/explains data

• Clustering: Learn the grouping

structure for a given set of
unlabeled inputs.

• Homogeneous groups as latent

structure: Clustering

• Other Applications: Topic

Modelling, Image Segmentation,
Social Networking.

62/92
 Unsupervised Learning: Dimensionality Reduction

• Given: Training data in

form of unlabeled instances
{x1 , x2 , ..., xN }

• Dimensionality Reduction:
Learn a Low-dimensional
representation for a given
set of high-dimensional
inputs

• Note: DR also comes in

supervised flavors
(supervised DR)

• Other Applications: facial

recognition, computer
vision and image
compression.

63/92
 Recipe for Learning

64/92
 T OPIC 4 : A RTIFICIAL INTELLIGENCE

65/92
 A first look at Artificial Intelligence

• What is Artificial Intelligence?

• What are the main challenges?

• What are the applications of AI?

• What are the issues raised by AI?

• On September 1955, a project

was proposed by McCarthy,
Marvin Minsky, Nathaniel
Rochester and Claude Shannon
introducing formally for the first
time the term "Artificial
Intelligence".

66/92
 Genesis of AI

67/92
 Misconception of AI

• AI is about electronic device able to mimic human thinking:

1 Artificial Intelligence.
2 One famous class of AI algorithms are called neural networks.
3 Android are close to humans in shape so they must think like human.

• Most AI algorithms do not aim at reproducing human reasoning.

• "The study and design of intelligent agents" where an intelligent agent

is a system that perceives its environment and takes actions that
maximize its chances of success - Frequent definition of AI.

• "In from three to eight years we will have a machine with the general
intelligence of an average human being." - Marvin Minsky (1970, Life
Magazine).

68/92
 AI is not human intelligence

"What often happens is that an engineer has an idea of how the brain
works (in his opinion) and then designs a machine that behaves that way.
This new machine may in fact work very well. But, I must warn you that
it does not tell us anything about how the brain actually works, nor is
it necessary to ever really know that, in order to make a computer very
capable. It is not necessary to understand the way birds flap their wings
and how the feathers are designed in order to make a flying machine [...]
It is therefore not necessary to imitate the behavior of Nature in
detail in order to engineer a device which can in many respects
surpass Nature’s abilities."
- Richard Feynman (1999).

69/92
 AI technology - Autonomous cars

• Originates from 1920 (NY)

• First use of neural networks to

control autonomous cars (1989)

• Four US states allow self-driving

cars (2013)

• First known fatal accident (May

2016)

• Singapore launched the first self-

driving taxi service (Aug. 2016)

• A Arizona pedestrian was killed

by an Uber self-driving car
(March 2018).

70/92
 AI technology - virtual assistant / chatbot

• Voice recognition tool "Harpy"

masters about 1000 words
(1970s, CMU, US Defense).

• System capable of analyzing

entire word sequences (1980).

• Siri was the first modern digital

virtual assistant installed on a
smartphone (2011).

• Watson won the TV show

Jeopardy! (2011).

71/92
 Exploring the beauty of pure mathematics

72/92
 Exploring the beauty of pure mathematics

• An Extract from a Post by DeepMind on December 1, 2021 (LINK):

- More than a century ago, Srinivasa Ramanujan shocked the mathematical

world with his extraordinary ability to see remarkable patterns in numbers that
no one else could see.
- These observations captured the tremendous beauty and sheer possibility of
the abstract world of pure mathematics.
- In recent years, we have begun to see AI make breakthroughs in areas
involving deep human intuition, and more recently on some of the hardest
problems across the sciences.
- As part of DeepMind’s mission to solve intelligence, we explored the potential
of machine learning (ML) to recognize mathematical structures and patterns,
and help guide mathematicians toward discoveries they may otherwise never
have found — demonstrating for the first time that AI can help at the forefront of
pure mathematics.

73/92
 T OPIC 5 : B IG D ATA

74/92
 Storage and Processing capacities
• Kryder’s Law is the assumption that disk drive density, also known as areal
density, will double every thirteen months. The implication of Kryder’s Law is
that as areal density improves, storage will become cheaper.

• Moore’s Law refers to Moore’s perception that the number of transistors on a

microchip doubles every two years. Moore’s Law states that we can expect the
speed and capability of our computers to increase every couple of years, and we
will pay less for them.

Storage capacity (Kryder’s law) Processor capacity (Moore’s law)

75/92
 How large your data is?

• What is the maximum file size

you have dealt so far?
(Movies/files/streaming video
that you have used)

• What is the maximum download

speed you get? (To retrieve data
stored in distant locations?)

• How fast your computation is?

(How much time to just transfer
from you, process and get
result?)

• “Every day, we create 2.5

quintillion bytes of data in 2020"
(So much that 90% of the data in
the world today has been created
in the last two years alone).
76/92
 Data Source: Examples

77/92
 Now data is Big data!
“Big data is data whose scale, diversity, and complexity require new architecture,
techniques, algorithms, and analytics to manage it and extract value and hidden
knowledge from it” - Standard definition.

Difficulties related to (Big) data:

• The prediction must be accurate: difficult for some tasks like image
classification, video captioning...
• The prediction must be quick: online recommendation should not take minutes.
• Data must be stored and accessible easily.
• It may be difficult to access all data at the same time. Data may come
sequentially.

78/92
 Characteristics of Big data: 3Vs

79/92
 Big data: V for Volume

Volume:
• Volume of data that
needs to be processed is
increasing rapidly.

• Need more storage

capacity.

• Need more
computation facility.

• Need more tools and

techniques.

80/92
 Big data: V for Variety

Variety:
• Various formats, types, and
structures.

• Text, numerical, images,

audio, video, sequences,
time series, social media
data, multi-dimensional
arrays, etc.

• A single application can be

generating or collecting
many types of data.

• To extract knowledge, all

these types of data need to
be linked together.

81/92
 Big data: V for Velocity

Velocity:
• Data is being generated fast
and need to be processed
fast.

• For time sensitive processes

such as catching fraud, big
data must be used as it
streams into your enterprise
in order to maximize its
value.

• Analyze 500 million daily

call detail records in
real-time to predict
customer churn faster.

• Scrutinize 5 million trade

events created each day to
identify potential fraud.

82/92
 Big data vs. Small data

Big data is more real time in nature than traditional applications

(Massively parallel processing, scale out architectures are well suited
for big data applications)...

83/92
 Challenges ahead. . .

• The Bottleneck is in technology:

New architecture, algorithms,
techniques are needed.

• Also in technical skills: Experts

in using the new technology and
dealing with Big data

• Who are the major players in the

world of Big data?

• Ethical issues: Tay ("thinking

about you" ) was an AI released
by Microsoft via Twitter in 2016.
It was shut down when the bot
began to post in inflammatory
and offensive tweets, only 16
hours after its launch.

84/92
 Big data landscape

85/92
 Still far from terminator

• Stephen Hawking BBC, Dec 2 2014

The development of full artificial intelligence could spell the end of the human
race. We cannot quite know what will happen if a machine exceeds our own
intelligence, so we can’t know if we’ll be infinitely helped by it, or ignored by it
and sidelined, or conceivably destroyed by it.

86/92
 Emerging Research Areas

Machine Learning: Big Data &

• Theoretical ML
Analytics: Other Areas:
(Explainable & Robust • Business Analytics • Quantum Computing
AI) (Quality, Reliability,
OR Modelling) • 3D Modelling, AR &
• Applied ML
VR
(Supervised & • Data Assimilation
Unsupervised • Blockchain &
Problems) • Methods for Big Data Cryptography
• Deep Learning • Robotics
(Application to CV, IP, • High-Dimensional
IOT, etc.) Statistics

87/92
 Now we are stepping into risk-sensitive
areas
Shifting from Performance Driven to Risk Sensitive (Ex-
plainable AI)...

88/92
 Take Home

• Emerging Research Areas in Statistics, Data Science & ML:

- All the areas of Machine Learning.
- High-Dimensional Small Sample Size Data.
- Spatio-temporal Data.
- Fairness in ML and Interpretable Deep Learning.
- Collaborative research with other disciplines.

• Statistical Thinking:
- Accepting randomness in formulating models and ideas.
- Realizing and Analyzing dependence of conclusions on assumptions.
- Measuring uncertainty in some ways without forgetting dependence
on assumptions.

89/92
 Textbooks

90/92
 Questions of the day

What type of data are involved in the following applications?

• Weather forecasting
• Mobile usage of all customers of a service provider
• Anomaly (e.g. fraud) detection in a bank organization
• Person categorization, that is, identifying a human
• Air traffic control in an airport
• Streaming data from all flying air crafts of Boeing

91/92
 End of Session

Caution: “Prediction is very difficult, especially if it’s about the future"

- Niels Bohr, Father of Quantum.

92/92