Introduction to Machine Learning

Course Logistics

Varun ChandolaComputer Science & Engineering

State University of New York at BuffaloBuffalo, NY, USA

chandola@buffalo.edu

Chandola@UB CSE 474/574 1 / 18

Outline

Class Details

Syllabus

Textbooks

Grading

Gradiance

Python

Socrative Online

Honor Code

Checklist and Resources

Warmup

Chandola@UB CSE 474/574 2 / 18

Class Details

I Lecture InformationI Monday, Wednesday, Friday (9.00 - 9.50 AM)I 109 Knox

I Recitations

1. 10.00 - 10.50 AM Monday, Norton 2102. 01.00 - 01.50 PM Tuesday, Bell 3373. 08.00 - 08.50 AM Friday, Cooke 127a

I Recitation topics are listed in the syllabusI No recitation this week.

I Class web pageI http:

//www.cse.buffalo.edu/~chandola/machinelearning.htmlI https://piazza.com/buffalo/spring2018/cse474/home

Chandola@UB CSE 474/574 3 / 18

Instructor

I Varun ChandolaI http://www.cse.buffalo.edu/~chandolaI Email: chandola@buffalo.eduI Office: 304 Davis HallI Phone: (716) 645-4747

I Office Hours: 1.00 PM - 3.00 PM (Mondays)

Chandola@UB CSE 474/574 4 / 18

Teaching Assistants

I Xin MaI Email: xma24@buffalo.eduI Office Hours: 10.00 AM - 11.00 AM (Fridays)

I Rudra Prasad BakshiI Email: rudrapra@buffalo.eduI Office Hours: 11.30 AM - 12.30 PM (Wednesdays)

I Hongfei XueI Email: hongfeix@buffalo.eduI Office Hours: 4.00 PM - 5.00 PM (Mondays)

Chandola@UB CSE 474/574 5 / 18

Piazza

I Primary medium of communication

I All announcements, teaching notes, slides, polls, etc. will be madeavailable through Piazza.

I Questions?1. General post to all (Name will be visible).

I Choose appropriate folder.

2. Private post to instructor, TA.

I Interact.

Piazza IncentiveI Top 3 contributors (questions or answers) will get recognized

I Award - To be decided

Chandola@UB CSE 474/574 6 / 18

Topics Covered

Theoretical Machine Learning

I Concept Learning

I Mistake Bound Online Learning

I Vapnik-Chervonenkis Dimension

I PAC Learning

I Statistical Learning Theory

Machine Learning Tools

I Bayesian Inference

I Expectation Maximization

I Optimization

Machine Learning Algorithms

I Linear Regression

I Linear Classification

I Neural Networks

I Support Vector Machines

I Kernel Methods

I Latent Space Models (PCA)

I Mixture of Models

I Bayesian Networks

Chandola@UB CSE 474/574 7 / 18

Topics Covered

Theoretical Machine Learning

I Concept Learning

I Mistake Bound Online Learning

I Vapnik-Chervonenkis Dimension

I PAC Learning

I Statistical Learning Theory

Machine Learning Tools

I Bayesian Inference

I Expectation Maximization

I Optimization

Machine Learning Algorithms

I Linear Regression

I Linear Classification

I Neural Networks

I Support Vector Machines

I Kernel Methods

I Latent Space Models (PCA)

I Mixture of Models

I Bayesian Networks

Chandola@UB CSE 474/574 7 / 18

Topics Covered

Theoretical Machine Learning

I Concept Learning

I Mistake Bound Online Learning

I Vapnik-Chervonenkis Dimension

I PAC Learning

I Statistical Learning Theory

Machine Learning Tools

I Bayesian Inference

I Expectation Maximization

I Optimization

Machine Learning Algorithms

I Linear Regression

I Linear Classification

I Neural Networks

I Support Vector Machines

I Kernel Methods

I Latent Space Models (PCA)

I Mixture of Models

I Bayesian Networks

Chandola@UB CSE 474/574 7 / 18

Topics Not Covered

I Deep learning

I Reinforcement Learning

I Probabilistic Graphical Models and Bayesian Networks

I Decision Trees

I Association Analysis

I Applications of Machine Learning (See Piazza post)

Chandola@UB CSE 474/574 8 / 18

Textbooks

I No prescribed text

I Primary references

I Optional reading list

Chandola@UB CSE 474/574 9 / 18

Grading

I Grading SchemeI Short weekly quizzes using Gradiance (12) – 20%I Programming Assignments (3) – 30%I Mid-term Exam (in-class, open book/notes) on 03/16/2018 – 20%I Final Exam (in-class, open book/notes) on 05/16/2018 – 30%

I All components will be individually curved

I Final grade:

A [92.5, 100]A- [87.5, 92.5)B+ [82.5, 87.5)B [77.5, 82.5)

B- [72.5, 77.5)C+ [67.5, 72.5)C [62.5, 67.5)C- [57.5, 62.5)

I Use UBLearns for all electronic submissions

Chandola@UB CSE 474/574 10 / 18

Mid-term and Final Exams

I All material covered in classI Final exam will not be comprehensive

I All multi-choice objective problems

I No partial credit

Chandola@UB CSE 474/574 11 / 18

Gradiance

I An online quiz system

I One quiz per week released on Monday by 8.59 AM and due nextSunday by 11.59 PM

I 3 - 4 multiple choice problems about topics covered that week

I A warm up quiz (ungraded) is posted

I 5-minute delay between successive submissions

I Only 3 tries allowed, maximum score will be used

I Every wrong answer will result in 1 negative point per try

Gradiance EnrollmentI Go to http://www.newgradiance.com/services

I Register and use the class token FC4761F5

I Make sure you register using your UBIT name as the username

I No other username will be accepted

Chandola@UB CSE 474/574 12 / 18

Python

I All programming assigments and class demonstrations using Python

I Resources:I Installing python, ipythonI Python IDE - CanopyI More about ipython notebooksI Python for Developers, a complete book on Python programming by

Ricardo DuarteI CodeAmerica - PythonI An introduction to machine learning with Python and scikit-learn

(repo and overview) by Hannes Schulz and Andreas Mueller

Github RepoI https://github.com/ubdsgroup/ubmlcourse

I http://nbviewer.ipython.org/github/ubdsgroup/

ubmlcourse/tree/master/notebooks/

Chandola@UB CSE 474/574 13 / 18

Socrative Online

I Online student response systemI Random number generator!

I http://m.socrative.com/student/

I Enter class ID - 259432

I Optional

Chandola@UB CSE 474/574 14 / 18

Academic Integrity and Honor Code

I http://www.cse.buffalo.edu/shared/policies/academic.php

Chandola@UB CSE 474/574 15 / 18

Machine Learning Honor Code

I Against the ML honor code to:

1. Collaborate on Gradiance quizzes2. Collaborate or cheat during exams3. Submit someone else’s work, including from the internet, as one’s

own for any submission4. Misuse Piazza forum

I You are allowed to:

1. Have discussions about homeworks. Every student should submitown homework with names of students in the discussion groupexplicitly mentioned.

2. Collaborate in groups of 2 or 3 for programming assignments. Onesubmission is required for each group.

I Violation of ML honor code and departmental policy will result in anautomatic F for the concerned submission

I Two violations ⇒ fail grade in the course

Chandola@UB CSE 474/574 16 / 18

Checklist and Resources

1. Sign-up for Piazza

2. Sign-up for Gradiance, try warm-up quiz

3. Read the department’s academic integrity policy

ResourcesI Piazza - piazza.com/buffalo/spring2018/cse474/home

I Video Channel - TBA

I Course slides and handouts -www.cse.buffalo.edu/~chandola/machinelearning.html

I Github Repo - github.com/ubdsgroup/ubmlcourse

I Notebooks - nbviewer.ipython.org/github/ubdsgroup/ubmlcourse/tree/master/

Chandola@UB CSE 474/574 17 / 18

Warmup Exercise

A fair coinI Probability of heads?

I 5 heads in a row?

I 5th head after seen 4 heads in arow?

I Gambler’s Fallacy

I If I know that probability of twopeople bringing a bomb on aplane is very low, should I bringa bomb along to make myselfsafer?

Matrix Vector ProductsI Let [3, 4] denote a vector in a

2D space

I Multiply with a number?

2

[34

]I Multiply with a matrix?[

2 1−2 3

] [34

]I For a matrix, find a vector such

that matrix-vector product ≡scalar-vector product.

Chandola@UB CSE 474/574 18 / 18

Warmup Exercise

A fair coinI Probability of heads?

I 5 heads in a row?

I 5th head after seen 4 heads in arow?

I Gambler’s Fallacy

I If I know that probability of twopeople bringing a bomb on aplane is very low, should I bringa bomb along to make myselfsafer?

Matrix Vector ProductsI Let [3, 4] denote a vector in a

2D space

I Multiply with a number?

2

[34

]I Multiply with a matrix?[

2 1−2 3

] [34

]I For a matrix, find a vector such

that matrix-vector product ≡scalar-vector product.

Chandola@UB CSE 474/574 18 / 18

Warmup Exercise

A fair coinI Probability of heads?

I 5 heads in a row?

I 5th head after seen 4 heads in arow?

I Gambler’s Fallacy

I If I know that probability of twopeople bringing a bomb on aplane is very low, should I bringa bomb along to make myselfsafer?

Matrix Vector ProductsI Let [3, 4] denote a vector in a

2D space

I Multiply with a number?

2

[34

]I Multiply with a matrix?[

2 1−2 3

] [34

]I For a matrix, find a vector such

that matrix-vector product ≡scalar-vector product.

Chandola@UB CSE 474/574 18 / 18

Warmup Exercise

A fair coinI Probability of heads?

I 5 heads in a row?

I 5th head after seen 4 heads in arow?

I Gambler’s Fallacy

I If I know that probability of twopeople bringing a bomb on aplane is very low, should I bringa bomb along to make myselfsafer?

Matrix Vector ProductsI Let [3, 4] denote a vector in a

2D space

I Multiply with a number?

2

[34

]I Multiply with a matrix?[

2 1−2 3

] [34

]I For a matrix, find a vector such

that matrix-vector product ≡scalar-vector product.

Chandola@UB CSE 474/574 18 / 18