CSCI 4155/6505 Machine Learning (and Robotics) 2010
... Instructor: Dr. Thomas Trappenberg ...
Office:Room 4216 (in Mona Campbell Building on Coburg RD) or 313 in Goldberg Building (after .... email: email@example.com
.... Office hour: most times I'm in my office or send email
Lecture notes: (will be regularly updated)
Sept 9 (original handout)
Sept 10 (small change in the last paragraph of intro, and now includes chapter 2, Matlab)
Sept 15 (now includes chapter 3, basic robotics with Lego NXT, and Appendix A with Matlab NXT commands)
Sept 23 (now includes chapter 4 on probability theory)
Oct 4 (now includes chapter 5 on regression, classification and maximum likelihood)
Oct 12 (small modivications at the end of chapter 5 to mention cross-validation and boosting. Now includes Discriminant analysis (chapter 6) and beginning of general learning machines (chapter 8).
Oct 18 (includes SVM section and small additions to section 5.4)
Oct 25 (includes chaper 10 on reinforcement learning)
Nov 2 (some update on reinforcement learning section to clarrify some of the algorithm [Thanks to Tomasz and Bartlomiej for some good observation])
Nov 16 (includes now chapter 7 on graphical models)
Nov 17 (fixed some typos in the chapter 7 and added one exercise)
Nov 23 (Includes unsupervised learning)
- Submit the Matlab files for the exercises of Section 2.4 as attachments by email to firstname.lastname@example.org with the subject line A1 by Friday, September 17.
- Submit a Matlab program for the exercise in Section 3.5.1 for your group as attachments by email to email@example.com with the subject line A2 by Thursday, September 30 before 4pm. To clarfify, the program should be able tomake a map of the state space with obstracles. The program should then be able to specify the starting point and end point of a path and executes a valid path between these points.
- This assignment is a combination of a group assignment and an individual assignment covering the exercises at the end of chapter 4. Please submit the question for the group with subject line A3.1 and questions 2 and three individually with subject line A3.2. The assignment is dueFriday, October 8.
- Please submit individual answers to the three exercise questions in section 5.8 by email to firstname.lastname@example.org with subjectline A4. Deadline is Thursday, October 14.
- A. An outstanding researcher at Dalhousie University has devised four measures to predict if students will plagiarize in any class. The first measure is a combination of huge values of hair, eye and shoe color, the second is a normalized measure of the length of the right toe, the third is a seat-to-behind rubbing factor when listening to lectures, and the fourth is classified. A data set with labels is provided, where the last column has the value zero if the student plagiariezed and the value one otherwise.
Your research group is asked to provide predictions for the unlabeled cases provided in the data set without labels. Use logistic regression, discriminant analysis, and a support vector machine to predict the cases. Send your programs to make these predictions to email@example.com with subject line `a4' by Thursday, Oct 21, before class. Include a short discussion of the different methods. Also provide your best prediction with your method of choice (either from the previous methods or another method you prefer) .
B. Submit a brief research proposal to firstname.lastname@example.org with subject line `proposal'.
- Please submit individual answers to the exercise in section 7.2 and the first question of the exercise at the end of chapter 7. Send your Matlab programs (or analytic derivation for exercise in section 7.2) by email to email@example.com with subjectline A6. Deadline is Thursday, Novermber 25.
- Project paper is due on Wednesday December 15, 2010. Please submit electronically to firstname.lastname@example.org. For grad students taking CSCI 6505, the project should be written in a journal/conference format with a page limit of 6 pages. Additional pages will be discarded. Don't forget proper references. For ugrad students taking CSCI 4155, you don't have a page limit and have a free format. But include your name and affiliation, an abstract, and references where necessary.
Functions creature_state.m and main.m for example in Chapter 2
A demonstration program of the A* search algorithm, written by Bob Sturm. A simplified matlab function based on this demonstration is AstarPathPlanner.m.
An often used implementation of SVMs with MATLAB interface is LIBSVM
Sone good references to further references to SVMs are:
- SMOLA, Alex J. and Bernhard SCHÖLKOPF, A Tutorial on Support Vector Regression, 1998.
- GUNN, S., Support Vector Machines for Classification and Regression, ISIS Technical Report, 1998.
- COLLOBERT, R and S BENGIO, SVMTorch: Support Vector Machines for Large-Scale Regression Problems, Journal of Machine Learning Research, 2001.
A brief guide to writing a scientific paper.
Simplified RBM letter example
Please familiarize yourself with the university policy on Intellectual Honesty. Every
suspected case will be reported.