Machine Learning 2

Course manual 2020/2021

Course content

This course continues where Machine Learning 1 stopped. We will treat chapters 2, 8, 9, 10, 11, 13 of the book 'Pattern Recognition and Machine Learning' by C. Bishop, plus some additional material made clear on Canvas. Topics include:

• Exponential families
• Conditional independence
• Information theory
• Independent components analysis
• Graphical models
• Latent variable models
• Learning, exact and approximate inference
• Variational Inference
• Sampling methods, MCMC, etc.
• Sequential data models
• Causality

Study materials

Literature

• C.M. Bishop: 'Pattern Recognition and Machine Learning'. Springer, 2006. ISBN 0-38-731073-8.

• D.J.C. MacKay: 'Information Theory, Inference and Learning Algorithms'. CUP, 2003.

• K.P. Murphy: 'Machine Learning: A Probabilistic Perspective'. MIT Press, 2012.

• D. Barber: 'Bayesian Reasoning and Machine Learning'. CUP, 2012.

Objectives

• The student will be able to write down the definitions of all taught concepts (information quantities, graphical models, inference methods, learning, etc.), explain their meanings, derive corresponding formulas and reason about/within these frameworks. They can explain the differences, pros and cons of those methods.
• The students will be able to apply all the taught concepts to complex real world problems and implement them in python.
• The student will be able use the taught methods to analyze complex data sets and evaluate the resulting models.

Teaching methods

• Lecture
• Computer lab session/practical training
• Self-study
• Seminar
• Working independently on e.g. a project or thesis

Lectures provide an overview of the material, reading the book helps understanding the details, practice/homework exercises and TA sessions let you test and develop your understanding of the theory and programming exercises develop your implementation skills.

Learning activities

Attendance

This programme does not have requirements concerning attendance (OER part B).

Assessment

There will be (ungraded) practice exercises/homework, which will be discussed and practiced during the TA sessions (2 times per week).

There will be 4 tests in total, one every 2 weeks during the last TA session, lasting for about 1 hour and covering the content of the lectures up to the week before and corresponding practice exercises. Each of the tests will contribute 25% to the final grade.

There will be lab assignments. Only a (random) selection of the problems will be graded (unknown to the students in advance). Those points will count as bonus. The average grade obtained will be added as a bonus (up to max 1pt) to the final grade (before rounding).

In case the student participates in the resit exam, the grade of the resit exam will replace all other grades to 100%. Also no bonus points apply.

Assignments

See above.
All assignments have to be made individually. Feedback can be obtained via the TAs.

Fraud and plagiarism

The 'Regulations governing fraud and plagiarism for UvA students' applies to this course. This will be monitored carefully. Upon suspicion of fraud or plagiarism the Examinations Board of the programme will be informed. For the 'Regulations governing fraud and plagiarism for UvA students' see: www.student.uva.nl

Course structure

Timetable

The schedule for this course is published on DataNose.

Additional information

Prior knowledge: It is required to have successfully finished "Machine Learning 1" before one can register for "Machine Learning 2". The student should also be familiar with linear algebra, probability theory, calculus and programming in python.

Contact information

Coordinator

• dr. P.D. Forré