UofC Statistical Learning Study Group

Public repository for the UofC Statistical Learning study group. The group met over the summer of 2019 for 12 presentations given by university students and staff on different topics. The objective of the group was to develop foundational knowledge and explore new developments in the fields of statistics and machine learning.

The group was sponsored by the GSA Quality Money Fund and The UofC Biomedical Engineering Graduate Student Commitee (BMEG).

Presenting Members:

• Jordan Bannister
• Deepthi Rajashekar
• Lucas Lo Vercio
• Anthony Winder
• Anup Tuladhar
• Luis Souto Maior
• Matthias Wilms
• Banafshe Felfeliyan
• Van Anh Le
• Sebastian Crites
• Nagesh Subbanna
• Bryce Besler

Meetings

We are planning to meet every second week (roughly) on Friday from 12-1pm beginning April 5. Food will be provided.

Course Outline

Date	Topic	Presenter	Room
Friday Apr. 5rd	Information Theory	Jordan	1405A
Friday Apr. 26th	Stochastic Processes	Deepthi	G732
Friday May. 3rd	Decision Trees/Forests	Lucas	G643
Friday May. 17th	Reinforcement Learning	Anthony	G643
Friday May. 31st	Generative Models	Anup	G643
Friday Jun. 7th	Interpretable Models	Luis	G732
Friday Jun. 14th	Computational Anatomy	Matthias	G384
Friday Jun. 21st	Recurrent Neural Networks	Banafshe	G382
Friday Jul. 12th	Ensemble Methods	Ahn	G382
Friday Jul. 26th	Geometric Deep Learning	Sebastian	G384
Friday Aug. 9th	Probabilistic Graphical Models	Nagesh	G384
Friday Aug. 23rd	Causal Inference	Bryce	G384

Curriculum

The content was selected according to the interests and knowledge of the presenting members. Related topics are (roughly) grouped together.

1 Information Theory

Information, entropy (conditional, joint, relative, differential), mutual information

2 Stochastic Processes

Random walk (levy process), brownian motion, gaussian process, markov process, martingale

3 Decision Trees/Forests

Decsion trees, bagging (bootstrap aggregation), random forests, information gain

4 Reinforcement Learning

Markov decision process, policy learning (brute force, monte carlo, Q-learning), exploration vs exploitation (multi-armed bandit problem)

5 Generative Models

VAE, GAN, deep belief network, style transfer.

6 Interpretable Models

Explainability vs interpretability; feature visualization; activation-based, gradient-based, and embedding-based methods.

7 Computational Anatomy

Diffeomorphisms (morphisms, isomorphisms, homeomorphisms, manifolds), diffeomorphism groups, matching/registration (LDMM)

8 Recurrent Neural Networks

Fully recurrent network, LSTM, training (supervised, reinforcement)

9 Ensemble Methods

Blending, bagging, boosting, stacking.

10 Geometric Deep Learning

Convolutions on graphs and manifolds, graph/manifold CNN’s, laplacian eigenbasis decomposition of graphs and manifolds.

11 Probabilistic Graphical Models

Bayesian networks, markov networks, conditional independence, joint probability factorization, hidden markov models

12 Causal Inference

Association, causation, intervention, counterfactuals, instrumentals, Structural Causal Models (SCM)