CS3390/CS5590/AI2000

Foundations of Machine Learning

Course Introduction
Aug 2024

Vineeth N Balasubramanian
 A few (not so, yet) recent quotes
• “A breakthrough in machine learning would be worth
ten Microsofts” (Bill Gates, Chairman, Microsoft)
• “Machine learning is the next Internet” (Tony Tether,
Director, DARPA)
• “Machine learning is today’s discontinuity” (Jerry
Yang, ex-CEO, Yahoo)
• “Machine learning is today’s electricity” (Andrew Ng)

Foundations of Machine Learning

What is Machine Learning?

Foundations of Machine Learning

 What is Machine Learning?
• Making predictions or decisions from data
• “Programming computers to optimize a performance criterion
using example data or past experience” (Ethem Alpaydin,
Machine Learning, 2010)
• “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.” (Tom Mitchell, Machine Learning, 1997)
• “Learning general models from a data of particular examples”
• “Build a model that is a good and useful approximation to the
data.“

Foundations of Machine Learning

 Today
Traditional Programming

Data
Computer Output
Program

Machine Learning
Data
Computer Program
Output
Source: Domingos

Foundations of Machine Learning

 Related Terms
Machine Learning, Data Mining, Knowledge Discovery,
Artificial Intelligence, Statistical Learning, Pattern Recognition,
Computational Learning

Foundations of Machine Learning

Applications of Machine Learning

Foundations of Machine Learning

 ML Applications
• Science (Astronomy, neuroscience, medical imaging, bio-informatics)
• Environment (energy, climate, weather, resources)
• Retail (Intelligent stock control, demographic store placement)
• Manufacturing (Intelligent control, automated monitoring, detection
methods)
• Security (Intelligent smoke alarms, fraud detection)
• Marketing (promotions, ...)
• Management (Scheduling, timetabling)
• Finance (credit scoring, risk analysis...)
• Web data (information retrieval, information extraction, ...)

Foundations of Machine Learning

 More Recent ML Applications
• Generative AI
• AlphaGo!
• Other unique ones:
• Automating Employee Access Control
• Identifying whales in ocean based on audio recordings
• Predict wait times for patients in emergency rooms
• Is (s)he a pyschopath?

Source: http://www.forbes.com/sites/85broads/2014/01/06/six-novel-machine-learning-applications/#6b6f9a9e67bf

Foundations of Machine Learning

 When is Machine Learning Used?
• Human expertise does not exist
• E.g. navigating on Mars
• Humans are unable to explain their expertise
• E.g. speech recognition
• Solution changes in time
• E.g. routing on a computer network
• Solution needs to be adapted to particular
cases
• E.g. user biometrics
• Data is cheap and abundant; knowledge is
expensive and scarce

Foundations of Machine Learning

 When are ML algorithms not needed?
• When the relationships between all system variables (input,
output, and hidden) is completely understood!

• This is NOT the case for almost any real system!

Foundations of Machine Learning

 Overview of ML
• Supervised learning
• Predict an output y when given an input x
• For categorical y : classification.
• For real-valued y : regression.
• Unsupervised learning
• Create an internal representation of the input, e.g. clustering, dimensionality
• This is important in machine learning as getting labels is often difficult and
expensive
• Other settings of ML
• Reinforcement learning (learning from “rewards”)
• Semi-supervised learning (combines supervised + unsupervised)
• Active learning, Transfer learning, Structured prediction

Foundations of Machine Learning

Classification (Supervised Learning)

Foundations of Machine Learning

Classification (Supervised Learning)

Foundations of Machine Learning

Regression (Supervised Learning)

Foundations of Machine Learning

Clustering (Unsupervised Learning)

Foundations of Machine Learning

 Dimensionality Reduction (Unsupervised Learning)
• Large sample size is required for high-dimensional data
• Query accuracy and efficiency degrade rapidly as the dimension
increases
• Strategies
• Feature reduction
• Feature selection
• Manifold learning
• Kernel learning

Foundations of Machine Learning

 Other Settings: Reinforcement Learning

• Learning a policy: A sequence of outputs

• No supervised output but delayed reward
• E.g. Game playing
• E.g. Robot in a maze
• Multiple agents, partial observability, ...

Foundations of Machine Learning

ML Problems

Foundations of Machine Learning

 ML in Practice
• Understanding domain, prior knowledge, and goals
• Data integration, selection, cleaning, pre-processing, etc.
• Learning models
• Interpreting results
• Consolidating and deploying discovered knowledge
• Loop

Foundations of Machine Learning

Training and Testing ML Models

Data acquisition Practical usage

Universal set
(unobserved)

Training set (observed) Testing set (unobserved)

Foundations of Machine Learning

 Training and Testing ML Models
• Training is the process of making the system able to learn.

• No free lunch rule:

• Training set and testing set may not come from the same distribution
• Need to make some assumptions or bias

Foundations of Machine Learning

 Types of Models
• Inductive vs Transductive Learning
• Online vs Offline Learning
• Generative vs Discriminative Models
• Parametric vs Non-Parametric Models

Foundations of Machine Learning

 Course Details
• Please see course web link

Foundations of Machine Learning

 Topics not to be covered (likely)
• Deep Learning (not in its “depth”, at least)
• Probabilistic Graphical Models/Bayesian Networks (not in
depth)
• Reinforcement Learning

Foundations of Machine Learning

 Mathematical Basis
• Functions, Logarithms and Exponentials
• Linear Algebra
• Vectors, Dot Products, Orthogonality
• Matrices, Matrix Operations, Linear Transformations,
Eigendecomposition
• Calculus, Differentiation, Integration (Vector and Matrix)
• Probability and Statistics
• (Optional; if you want theory)
• Real Analysis, Functional Analysis (Hilbert Spaces), Measure Theory,
Differential Geometry

Foundations of Machine Learning

 Programming
• Python
• Libraries
• Numpy, Scipy – numerical/scientific computing, linear algebra
• Matplotlib – for plotting
• Scikitlearn – for machine learning

Foundations of Machine Learning

 Resources to Learn & Practise ML
• MOOCs
• NPTEL, Coursera, EdX, Udacity
• Datasets
• UCI Repository, Statlib, Kaggle
• Conferences/Journals
• JMLR, TMLR, IEEE Transactions on Pattern Analysis and Machine
Intelligence, IEEE Transactions on Neural Networks and Learning
Systems
• ICML, NeurIPS, ICLR
• arXiv has the latest!

Foundations of Machine Learning

 Foundational Reading/Follow-up
• Math
• Part 1 of Deep Learning book: http://www.deeplearningbook.org/
• Essence of linear algebra: http://youtu.be/kjBOesZCoqc
• Essence of calculus: https://goo.gl/Hnk1jA
• Programming
• Learn Python
• https://try.jupyter.org/
• https://docs.python.org/3/tutorial/
• Video Tutorials: https://www.youtube.com/watch?v=cpPG0bKHYKc
• Play with Numpy

Foundations of Machine Learning