Beyond The Numbers

Beyond the Numbers
askokoad@gmail.com
1ZOGK3IHL5
Discovering the fascinating history

of Data Science
Proprietary content. © Great Learning. All Rights Reserved. Unauthorized use or distribution
This file is meant for personal use by askokoad@gmail.com only.
Sharing or publishing the contents in part or full is liable for legal action.
Proprietary content. © Great Learning. All Rights Reserved. Unauthorized use or distribution prohibited.
Meet Your Speaker
Dr. Abhinanda Sarkar

Academic Director at Great Learning
● Alumnus - Indian Statistical Institute, Stanford University

● Taught at - MIT, Indian Institute of Management, Indian Institute of
askokoad@gmail.com
1ZOGK3IHL5 Science
● Experienced in applying probabilistic models, statistical analysis and
machine learning to diverse areas
● Certified as Master Black Belt in Lean Six Sigma and Design
for Six Sigma in GE
Video Thumbnail
DO NOT REMOVE
Learning Objectives
Upon completion of this module, you will be able to:
Describe the historical context of decision-making.
Illustrate and differentiate between inferential and computational paradigms of data

science.
askokoad@gmail.com
1ZOGK3IHL5
Explore the evolution of inferential and computational paradigms over the years through
key innovations and their drivers.
Interpret the impact of these innovations on further advancements and developments.
Video Thumbnail
DO NOT REMOVE
Learning Outcomes
You should now be able to:
Comprehend what data science and analytics means.
Understand the idea of data-driven decision making.
askokoad@gmail.com
1ZOGK3IHL5 Illustrate how data science has evolved over the past century.
Identify the difference approaches within data science.
Video Thumbnail
DO NOT REMOVE
Agenda
In this session, we will discuss:
Origin of Decisions
Paradigms in Data Science
askokoad@gmail.com Evolution of Data Science (From the 1940s till now)

1ZOGK3IHL5
Video Thumbnail
DO NOT REMOVE
Origin of Decisions
Decisions were always data-driven.
Let’s consider a few situations that early civilizations might have faced:
When to stock up on firewood?

Geography
Growth & Sustenance of
the community Pressure
askokoad@gmail.com
What is the best time to travel?
1ZOGK3IHL5
Predict the weather ? Rainfall
inputs Temperature
Humidity
When to start cultivating a crop?
Decisions are made today by businesses in the same way - but the methods
have become more accurate and faster owing to the evolution of Video Thumbnail
statistical techniques & computing capabilities. DO NOT REMOVE
Paradigms in Data Science
Inferential Computational
Make predictions on population based on Leverage computational methods and

sample data technology to scale insight generation Focus
Use statistical methods to draw conclusions / Implement algorithms and computational

inferences from data methods to analyse data Methods
askokoad@gmail.com
1ZOGK3IHL5 Complexity of algorithms and cost of training
Representativeness of data Limitations
large models
1. Effectiveness of a new medication through 1. Weather forecasting based on historical and

randomized trial weather patterns
2. Impact of a new policy on citizens 2. Optimize routing of vehicles to minimize costs
Examples
Video Thumbnail
DO NOT REMOVE
Evolution of Data Science
askokoad@gmail.com
(Early Stat istics & Computing)

1ZOGK3IHL5
Video Thumbnail
DO NOT REMOVE
Inferential
Central Tendency
Expected Values
Probability Theory
< 1940s 1940s - 50s 1960s - 70s 1980s - 90s 2000s - 10s 2020+
askokoad@gmail.com
Telegraph
1ZOGK3IHL5
Difference Engine
Audio Tapes
Computational
Video Thumbnail
DO NOT REMOVE
Evolution of Data Science (1940s to 1950s)
Industrial Statistics
Use prior knowledge to predict future uncertainties
● Control Chart Theory - Dr. Shewhart @ Bell Labs

● Telephone manufacturing - Quality Control Process
● Led to economic benefits - High quality with efficiency
● Lower wastage + Higher product quality
Sampling Theory
askokoad@gmail.com
Make inferences about a population, using a sample
Inferential
1ZOGK3IHL5
● Frederick Taylor - Father of Scientific Management

● Manpower Productivity Assessments
● To improve manufacturing processes + Efficiency
● Now - Advertising by Google/Meta - Target audiences
ANOVA
Compare the means of 3+ groups - Evidence of difference
● Ronald A Fisher - Statistician & geneticist

Video Thumbnail
● Analyze experiments in Agriculture
● Effect of different fertilizers >> Differences in yields DO NOT REMOVE
● Now - Market research, finance,
This file is meant forquality control
personal use by askokoad@gmail.com only.
Proprietary content. © Great Learning. All Reserved.
Rights Reserved. Unauthorized use orprohibited.
distribution
Proprietary content. © Great Learning. All Rights Unauthorized use or distribution
Digital Computers
Electronic device which can do math and logical calculations
● Electronic Numerical Integrator & Computer (ENIAC)

● Developed in the US - World War II - Artillery Firing Tables
● Finding trajectories for different types of guns
● Type of ammo + external conditions (temp & wind)
Monte Carlo Methods

Computational
askokoad@gmail.com
1ZOGK3IHL5 Run simulations with random inputs to arrive at conclusions
● Origin - Monte Carlo Casino – Chance-based games

● Developed during the Manhattan Project - 1940s
● Simulate the behaviour of neutrons in a nuclear reactor
● Now - Predict weather patterns, financial markets
Programming Languages
To have a computer understand instructions & execute them
● Fortran – The 1st programming language created

Video Thumbnail
● Made for scientific and engineering calculations
● Led to the development of the World Wide Web DO NOT REMOVE
● Revolutionized communication & personal
This file is meant for learninguse by askokoad@gmail.com only.
Sharing or
Proprietary publishing
content. © Greatthe contents
Learning. in part
All Rights or fullUnauthorized
Reserved. is liable foruse
legal action.
or distribution
Inferential
Central Tendency Industrial Statistics
Expected Values Sampling Theory
Probability Theory ANOVA
< 1940s 1940s - 50s 1960s - 70s 1980s - 90s 2000s - 10s 2020+
askokoad@gmail.com
Telegraph Digital Computers
1ZOGK3IHL5
Monte Carlo Methods
Difference Engine
Programming
Audio Tapes Languages
Computational
Video Thumbnail
DO NOT REMOVE
Non-Parametric Methods
Rely on ranking/ordering of data rather than the distribution
● Frank Wilcoxon - Wilcoxon Rank-sum test

● Effect of store ambiance on customer behavior
● Sales difference b/w 2 groups in 2 different stores
● More effective marketing and pricing strategies
Decision Theory
askokoad@gmail.com
Assign probabilities to different outcomes to make a decision
Inferential
1ZOGK3IHL5
● Howard Raiffa - Economist - Negotiation Processes

● Government agencies - achieve favorable outcomes
● Identify optimal strategies using decision trees
● Improved negotiation - labor, international trade
Robust Statistics
Provide accurate results despite outliers/extreme values
● John W Tukey - Statistician - Contributed to EDA

Video Thumbnail
● Improve QC process - Identify and remove outliers
● Box plots - identify outliers and variation DO NOT REMOVE
● Manufacturing better products,
This file lesser
is meant for defects
Sharing content.
Proprietary or publishing
© Great the contents
Learning. in part
legal action.
or distribution
Operating Systems
A software that manages resources & apps in a computer
● General Motors are responsible for creating the 1st OS
● GM-NAA I/O - designed for their IBM 704 mainframe
● To manage hardware and use them efficiently
● Today - Windows, MAC, Android, and Linux are everywhere
Databases & Storage

Computational
askokoad@gmail.com
1ZOGK3IHL5 Store, Organize & Query large amounts of data quickly
● CharlesBachman
Charles Bachman - IBM
- IBM - 1st
- 1st DBMSDBMS ever
ever created
created
● Businesses wanted databases to be
Businesses wanted databases to be standardizedstandardized
● CommonBusiness
Common Business Oriented
Oriented Language
Language - COBOL
- COBOL
● Laid
Laid the
theroots
rootsfor
forthe
thecreation of MySQL
creation in 1995
of MySQL in 1995
Time Sharing Systems

Many users can access a computer at the same time!
● CompatibleTime-Sharing
Compatible Time-Sharing System
System (CTSS)
(CTSS) Video Thumbnail
● Createdby
Created byMIT
MITtoto access
access oneoneIBMIBM computer
computer
● The seed
The seedthought
thoughtforfor modern
modern networking
networking systemssystems DO NOT REMOVE
● Today’s cloud Proprietary
exists content. ©
because of Great
the Learning. All Rights Reserved. Unauthorized use or distribution
experiment!
Today’s cloud exists
Thisbecause
file is meantofforthe experiment
Inferential
Non-Parametric
Methods
Decision Theory
Probability Theory ANOVA
Robust Statistics
< 1940s 1940s - 50s 1960s - 70s 1980s - 90s 2000s - 10s 2020+
askokoad@gmail.com Operating Systems

Telegraph
1ZOGK3IHL5
Monte Carlo Methods Databases &
Difference Engine
Difference Engine
Storage
Programming
Audio
AudioTapes
Tapes Languages Time-sharing Systems
Computational
Video Thumbnail
DO NOT REMOVE
Resampling Methods
Simulate multiple datasets from original data for analysis
● Bradley Efron- Bootstrap technique

● Market Research survey – Calculate the uncertainty of data
● Estimate sampling distribution + hypothesis testing
● More accurate estimates + better decision making
Generalized Linear Models

askokoad@gmail.com
Analyze data where outcome is not normally distributed
Inferential
1ZOGK3IHL5
● John Nelder + Robert Wedderburn

● Insurance claim modeling - Identify risk exposure
● Model different types of response variables
● More accurate + flexible modeling of several data types
Model Selection Techniques

Selecting best mathematical model for a process
● Akaike - Akaike Information Criterion (AIC)

Video Thumbnail
● Demand forecasting - Accurate predictions
● AIC - Evaluate model - Fit, explainability, accuracy DO NOT REMOVE
● Better model selection,
This file improved
is meant for prediction
personal useaccuracy
by askokoad@gmail.com only.
distribution
Personal Computers
Small, lightweight, affordable - used by a single person
● Altair 8880 - 1st PC - 1975 – A company called MITS

● This was primitive - Apple II in 1977 made PCs popular
● Mostly used by hobbyists & technicians
● You could work & collaborate from anywhere
Object-Oriented Programming
Computational
askokoad@gmail.com
1ZOGK3IHL5 An abstract entity with its own set of properties & functions
● Popularized by C++ and Java

● Revolutionized software development
● Flexible, Modular, Reusable & Easy to maintain codes
● Browsers, apps, games - Impossible w/o OOP
Advanced Programming Languages

High level - data structures - loops - objects - conditions
● Earliest was Fortran - Created in the mid-50s

Video Thumbnail
● Laid the foundation for Java, Ruby, and Python
● Able to create complex applications - At scale DO NOT REMOVE
● Democratization This
of programming - Anybody
file is meant for personal use can Code!
by askokoad@gmail.com only.
Proprietary content. © Great Learning. AllReserved.
distribution
Inferential
Non-Parametric
Non-Parametric Resampling Methods
Central Tendency
Industrial Statistics Methods
Methods
Generalized Linear
Expected Values
Sampling Theory
Decision Theory
Decision Theory Models
Probability
ProbabilityTheory
Theory ANOVA
ANOVA Robust
RobustStatistics
Statistics Model Selection
< 1940s 1940s - 50s 1960s - 70s 1980s - 90s 2000s - 10s 2020+
askokoad@gmail.com Operating Systems Personal Computers

Telegraph
1ZOGK3IHL5
Monte Carlo Methods Databases & Object Oriented
Difference Engine
Difference Engine
Storage Programming
Programming
Audio
AudioTapes
Tapes Languages Time-sharing Systems Adv. Programming
Computational
Video Thumbnail
DO NOT REMOVE
Bayesian Networks & Graphical Models
Relationship b/w variables in a dataset using graphs
● Judea Pearl - Directed Acyclic Graphs - Turing award

● Healthcare - Chances of the patient having a disease
● Easily interpret complex relationships b/w variables
● Accurate diagnosis + improved outcomes, decisions
Causal Inference
askokoad@gmail.com
Is change in one variable changing the other?
Inferential
1ZOGK3IHL5
● Donald Rubin - Rubin Causal Model

● Education - Summer program on student outcomes
● Using outcomes to represent causal relationship
● Better understanding of variables + decision making
Open Science Movement

Making research accessible, collaborative, transparent
● Eli Lilly - Open Innovation Drug Discovery (OIDD)

Video Thumbnail
● Pharma - Develop new drugs and treatments
● Greater collaboration, transparency, reproducibility DO NOT REMOVE
● Faster drug development + therapies
distribution
Artificial Intelligence
Machines responding/doing tasks at human-level intelligence
● Alan Turing - Machine Intelligence - Imitation Game

● Frank Rosenblatt - “Built the Perceptron” - late 50s
● 2012: Geoffrey Hinton - “Deep Neural Networks”
● 2016: AlphaGo defeats Human Go Champion
Big Data
Computational
askokoad@gmail.com
1ZOGK3IHL5 Massive digital information generated every second
● Nutch Search Engine - Optimize speed of search

● Doug Cutting, 2005: Hadoop (son’s toy elephant)
● Paved the way for in-memory computing: Spark
● Big Data Analytics - Started with Google & Facebook
Cloud Computing
Computing power & resources for everyone, on-demand
● Coined by Eric Schmitt - 2006, Google CEO

Video Thumbnail
● 2006: Amazon Web Services, the 1st cloud provider
● Followed by Microsoft Azure and Google Cloud DO NOT REMOVE
● Grew due to the This
reduction in the
file is meant forcost of computing
Sharing content.
Proprietary or publishing the contents
© Great Learning. in part
legal action.
or distribution
Inferential
Non-Parametric Resampling Methods Bayesian Networks
Central Tendency
Methods
Generalized Linear Causal Inference
Expected Values
Decision Theory Models
Probability Open Science
ProbabilityTheory
Theory ANOVA
Robust Statistics Model Selection Movement
< 1940s 1940s - 50s 1960s - 70s 1980s - 90s 2000s - 10s 2020s+
askokoad@gmail.com Operating Systems Personal Computers

Telegraph Digital Computers Artificial Intelligence
1ZOGK3IHL5
Monte Carlo Methods Databases & Object Oriented
Difference Engine Big Data
Storage Programming
Programming
Audio Tapes Languages Cloud Computing
Time-sharing Systems Adv. Programming
Computational
Video Thumbnail
DO NOT REMOVE
Evolution of Data Science (2020s+)
Interdisciplinary Approaches
Knowledge from multiple disciplines for problem solving
● Tesla - Advances in battery + electric motor tech
● Model S - Range of 400 km in a single charge
● Accelerated transition from fossil fuels
● Innovative solutions to complex problems
Newer Causal Inference methods

askokoad@gmail.com
Making causal inference accurate and reliable
Inferential
1ZOGK3IHL5
● Amazon - Personalized Marketing Campaigns

● Recommend products likely to be purchased
● Causal Inference methods - analyse user behavior
● Increased sales, customer satisfaction
Natural Experiments
Observe events naturally occuring w/o manipulating factors
● Journal of Public Economics - Study of policy impact
Video Thumbnail
● Effectiveness of Public health interventions
● Impact of business closure due to the pandemic on jobs DO NOT REMOVE
● Investigate complex phenomena
This file is meant for + precise
personal conclusions
use by askokoad@gmail.com only.
Proprietary content. © Great Learning. AllReserved.
Rights Reserved. Unauthorized use or distribution
Proprietary content. © Great Learning. All Rights Unauthorized use or distribution prohibitd.
Evolution of Data Science (2020s+)
Blockchain
Share information - secure, transparent, & tamper-proof
● Created for the proposal of a Virtual Currency System

● 2008: BitCoin - Underlying tech was Blockchain
● The concept is a threat for Traditional Banking Systems
● Extreme Security + Low Fees (No Central Authority)
Edge Computing
Computational
askokoad@gmail.com
1ZOGK3IHL5 Compute directly at the source of data, instead of remote
● Took shape in the early 2000s - Internet of Things (IoT)

● Tesla’s advancements with Autonomous Vehicles
● Opportunities: Real time monitoring & analysis (Medical Price of Bitcoin in USD
Devices, Defence, Smart Homes)
Quantum Computing
Use the principles of quantum physics to compute
● 1st built in 1998 - Los Alamos Laboratory New Mexico Video Thumbnail
● Impact areas: Cryptography, Chemistry & Optimization
● In early stages, a lot of opportunities are still theoretical and DO NOT REMOVE
under experimentationThis file is meant for personal use by askokoad@gmail.com only.
Proprietary
Sharing content. © Great Learning.
or publishing All Rights
the contents Reserved.
in part or fullUnauthorized
is liable foruse or distribution
legal action.
Inferential
Non-Parametric Resampling Methods Bayesian Networks Interdisciplinary
Methods Approaches
Generalized Linear Causal Inference
Decision Theory Models Newer Causal Inference
Open Science
Probability Theory ANOVA Natural Experiments
Robust Statistics Model Selection Movement
< 1940s 1940s - 50s 1960s - 70s 1980s - 90s 2000s - 10s 2020s+
askokoad@gmail.com Operating Systems Personal Computers Blockchain

Telegraph Digital Computers Artificial Intelligence
1ZOGK3IHL5
Monte Carlo Methods Databases & Object Oriented Edge Computing
Difference Engine Big Data
Storage Programming
Programming Quantum
Audio Tapes Languages Cloud Computing
Time-sharing Systems Adv. Programming Computing
Computational
Video Thumbnail
DO NOT REMOVE
Let’s conclude by defining data science
Machine
Learning &
Computer Deep Learning
Computer Science & Math & Statistics
Science
Technology
&
Technology
askokoad@gmail.com Data
1ZOGK3IHL5 Science
Software
Development Research
Domain/Business
Knowledge
Video Thumbnail
DO NOT REMOVE
Summary
Upon completion of this module, you will be able to:
Early civilizations made use of decisions to solve critical problems that impacted the growth
and sustenance of the community.
Data Science involves the amalgamation of two paradigms - the inferential and
computational paradigms.
The Inferential paradigm focuses on the statistical methods for analysis, while the
askokoad@gmail.com
1ZOGK3IHL5 computational paradigm focuses on computational methods and algorithms.
Data Science has evolved from simple probabilistic models and primary computers in the
early 1950s to highly advanced inference methods and computing in the 2020s.
The developments in inferential and computational paradigms act as catalysts that propel
further advancements in these categories as the evolution continues.
Video Thumbnail
DO NOT REMOVE
Ha p p y Lea rning !
askokoad@gmail.com
1ZOGK3IHL5
Sharing or publishing the contents in part or full is liable for legal action. 27

Beyond The Numbers - Discovering The Fascinating History of Data Science

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Beyond The Numbers - Discovering The Fascinating History of Data Science

Uploaded by

Copyright:

Available Formats

Discovering the fascinating history

Dr. Abhinanda Sarkar

● Alumnus - Indian Statistical Institute, Stanford University

Upon completion of this module, you will be able to:

Describe the historical context of decision-making.

Illustrate and differentiate between inferential and computational paradigms of data

Interpret the impact of these innovations on further advancements and developments.

You should now be able to:

Comprehend what data science and analytics means.

Understand the idea of data-driven decision making.

Identify the difference approaches within data science.

In this session, we will discuss:

Paradigms in Data Science

askokoad@gmail.com Evolution of Data Science (From the 1940s till now)

Decisions were always data-driven.

When to stock up on firewood?

Make predictions on population based on Leverage computational methods and

Use statistical methods to draw conclusions / Implement algorithms and computational

1. Effectiveness of a new medication through 1. Weather forecasting based on historical and

(Early Stat istics & Computing)

● Control Chart Theory - Dr. Shewhart @ Bell Labs

● Frederick Taylor - Father of Scientific Management

● Ronald A Fisher - Statistician & geneticist

● Electronic Numerical Integrator & Computer (ENIAC)

Monte Carlo Methods

● Origin - Monte Carlo Casino – Chance-based games

● Fortran – The 1st programming language created

Central Tendency Industrial Statistics

Expected Values Sampling Theory

Probability Theory ANOVA

● Frank Wilcoxon - Wilcoxon Rank-sum test

● Howard Raiffa - Economist - Negotiation Processes

● John W Tukey - Statistician - Contributed to EDA

Databases & Storage

Time Sharing Systems

askokoad@gmail.com Operating Systems

● Bradley Efron- Bootstrap technique

Generalized Linear Models

● John Nelder + Robert Wedderburn

Model Selection Techniques

● Akaike - Akaike Information Criterion (AIC)

● Altair 8880 - 1st PC - 1975 – A company called MITS

● Popularized by C++ and Java

Advanced Programming Languages

● Earliest was Fortran - Created in the mid-50s

askokoad@gmail.com Operating Systems Personal Computers

● Judea Pearl - Directed Acyclic Graphs - Turing award

● Donald Rubin - Rubin Causal Model

Open Science Movement

● Eli Lilly - Open Innovation Drug Discovery (OIDD)

● Alan Turing - Machine Intelligence - Imitation Game

● Nutch Search Engine - Optimize speed of search

● Coined by Eric Schmitt - 2006, Google CEO

askokoad@gmail.com Operating Systems Personal Computers

Newer Causal Inference methods

● Amazon - Personalized Marketing Campaigns

● Created for the proposal of a Virtual Currency System

● Took shape in the early 2000s - Internet of Things (IoT)

askokoad@gmail.com Operating Systems Personal Computers Blockchain

You might also like