Advanced Machine Learning (GR5242) • Fall 2018

This class continues from where Statistical Machine Learning (GR5241) left off. It covers neural networks, graphical models, sampling algorithms, and related topics.

Prerequisites

You must have attended Statistical Computation and Introduction to Data Science (GR5206) and Statistical Machine Learning (GR5241) in order to take this class. Please make sure you meet the prerequisites; if you enroll but do not have the prerequisites, the school will remove you from the class.

Syllabus

The syllabus for the class is available here.

Final projects

The list of possible final projects is available here.

Instructors

Peter Orbanz (first half of the term)
John Cunningham (second half of the term)

Office hours: After class in the class room

Teaching Assistants

Ian Kinsella (iak2119@columbia.edu)
Andrew Davison (ad3395@columbia.edu)
Gabriel Loaiza-Ganem (gl2480@columbia.edu)
Peter Lee (jl4303@columbia.edu)

Office hours: Monday, Tuesday, 5.30-7:00pm. Room 1025, Department of Statistics

Email Policy

All questions pertaining to the course (organization, curriculum, material, ect.) should be asked on Piazza where other students will be able to benefit from the answers. Questions containing sensitive &/or personal information can be directed to the TA’s who will redirect them to the professors as appropriate. No emails should be sent directly to the professors.

Piazza Policy

The TA’s will manage a Piazza forum for students to ask, and help answer, questions pertaining to course organization and material. Students are highly encouraged to browse previously answered questions and contribute answers to those still awaiting responses. Questions must be specific, self contained, and answerable with a single response. For example, a question troubleshooting the performance of a model needs to provide details about the model architecture, training procedure, dataset, and performance metrics. The TA’s will not be responsible for posting back and forth on the forums to extract these details.

If your question does not fit well within these confines, it should be brought to office hours where that type of discussion is more appropriate. Furthermore, the Piazza forums and office hours are specifically intended for discussion about the models and concepts covered in the course, NOT general Python or Tensorflow programming/debugging questions. Questions pertaining to these topics, in particular (but not limited to) “Why do I get this error message when I run my code?”, should be directed to appropriate online resources such as Stack Overflow.

Slides

  • Slides for Part I (last updated: 16th October)
  • Slides for Part II (last updated: 9th December)

Homework + useful files

Useful links

  • Course repository (containing Python/TF tutorials, code examples etc.)
  • Piazza (for questions relating to the course; note the Piazza policy above)

Textbooks

The course is not based on a textbook. The relevant course materials are the slides. If you would like to complement lectures and slides by further reading, these books might be useful:
  • Pattern Recognition and Machine Learning.
    Christopher M. Bishop.
    Springer, 2006.
  • Machine Learning: A Probabilistic Perspective.
    Kevin P. Murphy.
    MIT Press, 2012.
    [Available online]