Department of Computer Science and Technology

Course pages 2017–18

Probabilistic Machine Learning

Principal lecturer: Prof Carl Rasmussen
Additional lecturers: Prof Alan Blackwell, Dr Damon Wischik, Dr Maria Perez-Ortiz
Taken by: MPhil ACS, Part III
Code: LE49
Hours: 16 (14 lectures at the Department of Engineering, 2 examples classes; Lectures for this module are offered by the Department of Engineering, Trumpington Street site, during Michaelmas Term. Practical work in Lent term is administered and assessed by the Computer Laboratory.)
Class limit: 20 students
Prerequisites: A good background in statistics, calculus and linear algebra will be required. Note that the lectures in this course are primarily oriented toward Cambridge Engineering students. More detail of prerequisites should be reviewed on the Engineering Department website: http://mlg.eng.cam.ac.uk/teaching/4f13/1617/ Students are asked to talk with their Course Adviser before selecting this module.

Further information

This module's lectures are taught at the Department of Engineering.

This module is borrowed from the Department of Engineering and the lectures are given at the Trumpington Street site during Michaelmas Term. Students wishing to take this module should note that the Department of Engineering is about a 2 mile cycle ride from the Computer Laboratory via Adams Road or fifteen minutes by the Uni4 bus. See http://www.eng.cam.ac.uk/visitors/ for travel information. Assessment is carried out by the Department of Computer Science and Technology and continues in Lent Term.

Syllabus

Please refer to the Probabilistic Machine Learning syllabus for details of the material to be covered in lectures.

Practical work

Two practical challenges, defined in collaboration with professional machine learning researchers, and assessed by the Computer Laboratory, will be undertaken in Michaelmas and Lent Term. A report on the first challenge will be submitted at the end of Michaelmas term, and the second at the middle of Lent term.

A small selection of challenge topics will be advertised, drawing from the main themes of the lecture course:

  1. Application of Gaussian process models
  2. Probabilistic ranking applications
  3. LDA topic models

Assessment

  • Practical report 1 (easy challenge): 20%
  • Practical report 2 (hard challenge): 80%

Reports should be prepared in the style of a submission to the NIPS or ICML conference. The report on the easy challenge must be less than 4 pages. The report on the hard challenge must comply with the specification for a full paper at those conferences (8 pages including figures and tables, with an additional one or two pages allowed for references that have been cited).