| Instructor | Aaron Hertzmann | |
| hertzman@dgp.toronto.edu | ||
| Phone | (416) 946-8497 | |
| Office Hours | BA 5268, Wednesday 2-3pm (other times by appointment only) |
|
| Lectures | Monday/Wednesday 3-4pm (RW 229) | |
| Tutorials | Friday 3-4pm (BA 3012) | |
| Online | www.cs.toronto.edu/~csc411h |
This course introduces methods for automated learning of relationships on the basis of empirical data. Topics include classification and regression, nearest-neighbor methods, decision trees, linear models, neural networks, clustering algorithms, and Bayesian methods. Problems of overfitting, assessing accuracy, and handling large databases will be discussed.
The student is expected to be comfortable with basic probability and statistics, and coding. Assignments will be done in MATLAB/Octave or R; MATLAB and R are both available on CDF. However, the TAs and I are not experts in R, and may not be able to help much with programming questions; we recommend MATLAB or Octave for this reason.
There is no course textbook. Lecture notes will be provided.
| Assignment 1: | 15% |
| Assignment 2: | 20% |
| Assignment 3: | 20% |
| Midterm test: | 15% |
| Examination: | 30% |
As a general rule, small matters of marking on assignments (apparent addition errors, questions about evaluation criteria, etc.) should be taken first to the marker (via email). More significant issues, or unresolved matters on assignments, are appropriate to take to the professor. Matters of marking on tests and exams should be taken to the professor.
Assignments involve both theoretical problems as well as implementation
of algorithms.
Late assignments will be penalized 10% of the available
marks per day up to a maximum of three days; assignments will not be accepted after three days. No extensions
will be granted on homework assignments, except in extreme cases
(e.g.
medical reasons). Please plan ahead.
Plagiarism - or simply, cheating - is taken to be the handing in of work not substantially the student's own; it is usually done without reference, but is unacceptable even in the guise of acknowledged copying. It is reprehensible, and the penalty will be severe.
It is not cheating, however, to discuss ideas and approaches to a problem, nor is it cheating to seek or accept help with a program or with writing a paper. Indeed, a moderate form of collaboration is encouraged as a useful part of any educational process. However, good judgement must be used, and students are expected to present the results of their own thinking and writing. Never copy another students work -- it is plagiarism to do so, even if the other student "explains it to you first." Do not work together to form a collective solution, from which the members of the group copy out the final solution. Rather, walk away and recreate your own solution later. Note that it is also wrong to give work to other so that it may be plagiarised. Under no circumstances should give copies of your written work to others.
Remember, plagiarism is taken to be the handing in of work not substantially the student's own, and the penalty will be severe. If you have exchanged ideas with a fellow student and thus have answers which might be falsely construed as being plagiarised, you should clearly state this.
Please be courteous and professional in all electronic communications. Include your full name, either in the message body or the "From:" line. Anonymous emails and bulletin board postings may be ignored (or mistaken for spam). Including your CDF account name and/or student number in emails may help speed things up.
I recommend using the course bulletin board for discussion about class topics and homework. I will try to respond to emails sent directly to me within a few days.