SeminarHome

SEMINAR ANNOUNCEMENT - SPRING 2004

  ELEN E9701 - Seminar in Machine Learning and Signal Processing
  Fridays 10:30-12:00  7LE3 Schapiro CEPSR (Advent Seminar Room)
  Organizers: Prof. Dan Ellis <dpwe@ee.columbia.edu>

This 1-credit seminar class will cover a range of topics at the interface between modern techniques of machine learning and current problems in signal processing and interpretation. This semester we will be studying two main topics:

• Kernel methods in representation and learning - SVM and relatives
• Maximum entropy modeling and related information-theory concepts

Each week will consist of a brief reading, a presentation by one of the seminar participants, and a discussion. We will also be developing a "Wiki" web site (see the HomePage) including annotated references and demonstrations throughout the semester.

This class grows out of an existing reading group held by the students in the Advent (signal processing) group in the Electrical Engineering department, but we wish to make it open to any suitably interested student. However, a basic level of familiarity with topics in probability, machine learning, and signal processing, will be required, and in order to preserve the seminar atmosphere, we will have to limit total enrollment to 15.

For more information, see the TopicIdeas list.

For the developing schedule of the first half of the semester, see KernelMethods.

The second half of the semester will look at maximum entropy and other applications of information theory to machine learning; see InfoTheoryML.

MlspFall2004 home page.