Learning theory
Aperçu des semaines
-
This master class on learning theory covers the classical PAC framework for learning, stochastic gradient descent (together with recent research papers), tensor methods and/or graphical model learning.
Teachers: Ruediger Urbanke: ruediger.urbanke@epfl.ch and Nicolas Macris: nicolas.macris@epfl.ch
Teaching Assitant: Farzad Pourkamali: farzad.pourkamali@epfl.ch
Student Assistant: Yulun Jiang: yulun.jiang@epfl.ch
Courses: Mondays 8h15-10h in presence Room INM202; Exercises: Tuesdays 17h15-19h in presence Room INR219.
We will use this moodle page to distribute homeworks, solutions and also collect graded ones. As well as use the discussion and questions forum. Dont hesitate to actively use this forum.
Other information and notes material about the class is found on the web page: learning theory
If you miss a lecture a recorded version will be accessible here https://tube.switch.ch/channels/868590a6
-
If you have a question or want to start a discussion on a topic, post here
-
Chapter 3 in UML
Homework 1: exercises 1, 3, 7, 8 of Chapter 3.
-
Chapters 4 and 5 in UML
Homework 2: exercises 1 and 2 of chapter 4
-
Chapter 6 in UML
Graded hmw
extra optional: exercise 1 of chapter 5
-
Chapter 6 continued
graded hmw 3 continued
-
Chapter 7 in UML
Homework 4: exercise 3 of chapter 6 and exercise 3 of chapter 7.
-
Chapter 14 in UML: Gradient descent and stochastic gradient descent
-
Chapter 14 continued: Gradient descent and stochastic gradient descent
Class is by Zoom due to covid. Here is the link:
https://epfl.zoom.us/j/65366335305
-
Mean field approach for two layer neural networks
graded hmw 6 continued
-
Easter Week Break
-
Tensors 1. Motivations and examples, multi-dimensional arrays, tensor product, tensor rank.
-
Tensors 2. Tensor decompositions and rank, Jennrich's theorem
-
Tensors 3. Matricizations and Alternating Least Squares algorithm
-
Tensors 4. Multilinear rank Tucker higher order singular value decomposition
-
Tensors 5. Power method and Applications: Gaussian Mixture Models, Topic models of documents
-
End of power method; Review of class; Q&A.
-
Here are old exams with solutions. Note that most, if not all, problems have been already included in this year's material. In exam 2019 ignore problems on graphical models that we did not treat this year.