Scribd
Upload
60 views3 pages

CS5560: Probabilistic Models For ML

This course covers probabilistic models for machine learning, focusing on key techniques like maximum likelihood estimation, Bayesian estimation, linear and logistic regression, mixture models, hidden Markov models, and probabilistic graphical models. Students will learn these concepts through lectures, weekly assignments, and exams evaluating probabilistic modeling skills and mathematical understanding of machine learning. The textbook is "Machine Learning: A Probabilistic Perspective" and the instructor is available for office hours questions.

Uploaded by

Nishith Agarwal
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
60 views3 pages

CS5560: Probabilistic Models For ML

This course covers probabilistic models for machine learning, focusing on key techniques like maximum likelihood estimation, Bayesian estimation, linear and logistic regression, mixture models, hidden Markov models, and probabilistic graphical models. Students will learn these concepts through lectures, weekly assignments, and exams evaluating probabilistic modeling skills and mathematical understanding of machine learning. The textbook is "Machine Learning: A Probabilistic Perspective" and the instructor is available for office hours questions.

Uploaded by

Nishith Agarwal
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

CS5560: Probabilistic Models for ML

Overview
This is a foundational course in machine learning with focus on probabilistic
models. The course is intended for those who wish to pursue research in ma-
chine learning and related fields. Hence, the developments are technical and
involve appropriate mathematical details. Broadly, the course includes study-
ing basic probabilistic models that are often used in machine learning set-ups,
and methods of estimating parameters of such models from data. The course
ends with an introduction to a very rich class models known as probabilistic
graphical models1 .

Pre-requisites
The course assumes basic engineering level knowledge of probability theory,
and multivariate calculus. Familiarity with machine learning, statistics, and
mathematical optimization, will aid better appreciation of the course, but are
not necessary pre-requisites.

Syllabus and Text


The textbook for this course is “Machine Learning: A Probabilistic Perspective”
By Kevin P. Murphy (MIT Press). The detailed syllabus is:
1. Maximum likelihood estimation (MLE) in Binomial, Multinomial, Gaus-
sian, models in exponential family.
2. MAP, Bayesian estimation in Beta-Binomial, Dirichlet-Multinomial, Normal-
Inverse-Gamma-Gaussian, conjugate prior based models in exponential
family.
3. Learning with Supervised Models: Generative and discriminant models,
Linear Regression, Logistic Regression, Gaussian Discriminant Analysis,
Generalized linear models.
1 In the version of the course taught to EMDS/MDS students this advanced topic on Graph-

ical Models may not be covered.

1
4. Expectation Maximization (EM) based learning in Mixture models, Hid-
den Markov Model, Dirichlet processes (Clustering).
5. Introduction to directed (Bayes nets) and un-directed (Markov Random
Fields) graphical models.

Evaluation Scheme
Date Duration Percentage
31-08-2018 (Fri) 60 min. 15%
23-10-2018 (Tue) 90 min. 35%
23-11-2018 (Fri) 180 min. 50%

In addition to the above, there will be weekly assignments. In case these


are not submitted as per the instructions, two negative marks will be awarded
(per default). Some assignments will carry bonus marks and will be used for
grade-promotions during the final grading.
In case you are auditing this course, then 100% attendance is the only re-
quirement for passing the course.
For EMDS/MDS students all instructions remain same, except that there
will be only one exam (the one on the last row of the above table) carrying
100% marks.

Contact
You are welcome to drop-by my office (C-519) anytime during my regular of-
fice hours (8:30am-12:30pm;2:30pm-5:00pm). Though you are not required to
take prior permission, in case you don’t notify me, I may have stepped out for
some meeting/break when you visit. You are highly encouraged to ask ques-
tions/clarifications either in lecture or when you visit my office.
All correspondences in this course (including assignment submissions) will
be done via Google-Classroom (code will be announced in lecture). This forum
can also be used for clarifications/questions.

References by Topic
Here is list of sections in your textbook (Murphy’s book) for each topic covered
in the lectures:

Method of Moments: Simple method with no elaborate reference in your


textbook. Defined in equation (9.47).
Maximum Likelihood: Defined in equation (3.7). Some details in section
6.2.2.

2
Gaussian-MLE: Section 4.1.3.
Multinoulli-MLE: Again, easy! So not explicitly covered. Covered as special
case in section 3.2.

Generative-LinearRegression: Section 4.3.


Discriminative-LinearRegression: Sections 7.1-7.3.
GDA (BayesClassifier): Section 4.2.
Logistic Regression: Sections 8.1-8.3.

Exponential Family: Sections 9.1-9.2.


GLMs: Section 9.3.
Bayesian Inference and MAP: Sections 3.2.2-3.2.4, 3.3.2-3.3.4, 3.4.2-3.4.4,
3.5.1.2, 4.6, 7.5.1, 7.6 (non-*), 8.4 (light reading), 9.2.5.

Hierarchical Bayes: Sections 5.5, 5.6.


CrossValidation: Section 7.10 in https://web.stanford.edu/ hastie/Papers/ESLII.pdf.
Why it works is because of Lemma1 in www.kyb.mpg.de/publications/pdfs/pdf1436.pdf.
Model Selection (Bayesian): Sections 5.3, 5.5, 5.6.

Mixture Models: Section 11.2, 11.3.


EM for MixtureModels: Sections 11.4.1-11.4.2.
Markov Models: Section 17.2.

HMMs: Sections 17.3, 17.4, 17.5.


Bayes Nets: Entire chapter 10 except * sections. Additional reading: Chapter
3 except sections 3.3.3, 3.4.2, 3.4.3 in DaphneKoller’s book.
Markov Nets: Sections 19.1-19.4. Additional reading: CHapter 4 except sec-
tions 4.4, 4.5, 4.6 in DaphneKoller’s book.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy