EHB 328E - Machine Learning for Signal Processing

Course Objectives

(1) Introduce the student to the basic techniques of machine learning geared towards
signal processing applications
(2) Provide the students with non-trivial examples to develop their programming skills
(3) Gain a working knowledge of probability, linear algebra, signal processing
motivated by problems of current interest

Course Description

The course will include the following topics: Data-driven representations. Principal
Component Analysis (PCA) and Kernel PCA. Independent Component Analysis (ICA).
Non-negative matrix factorization (NMF). Dictionary based, sparse and overcomplete
data representations. Low rank matrix representations. Regression and Linear
prediction. Stochastic Gradient Descent and LMS adaptive filters. Clustering and
Classification. Neural Networks. Convolutional networks and applications to signal and
image processing. A good knowledge of probability theory, linear algebra and signals
and systems theory is a prerequisite for the course. The term project and homework
will necessitate software simulations.