courses:be4m33ssu:micro-credentials:discriminative-learning

Foundations of discriminative learning and empirical risk minimisation

Foundations of discriminative learning and empirical risk minimisation

Overview

This teaching block provides fundamentals of discriminative learning and covers generalisation error bounds, empirical risk minimisation (ERM), and its consistency. It introduces the class of structured output SVMs and shows how to apply ERM based learning for models in this class.

Teacher: Dr. Vojtech Franc

Prerequisites:
- probability theory and statistics
- linear algebra and optimisation
- pattern recognition and decision theory

Lectures

Topic	Recording	Notes
1. Empirical risk	mp4	[1] Chap 2, [2] Chap 7
2. Empirical risk minimization	mp4	[1] Chap 2, [2] Chap 7
3. Empirical risk minimization II	mp4	[1] Chap 4, [2] Chap 12
4. Structured Output Support Vector Machines	mp4	[1] Chap 5, [2] Chap 12

Theoretical assignments

Topic	Pdf
Seminar: lecture 1
Seminar: lecture 1,2
Seminar: lecture 2,3

Homework

Structured Output Perceptron task

Textbooks and References

[1] M. Mohri, A. Rostamizadeh and A. Talwalkar, Foundations of Machine Learning, MIT Press, 2012 [PDF]
[2] T. Hastie, R. Tibshirani and J. Friedman, The Elements of Statistical Learning, Springer, 2010 [PDF]

Table of Contents