CS 463 Projects

Semester grades will be based on assignments, which will include programming from time to time, and on each student's choice of two larger assignments. The "larger assignments" can be

programming projects;
presentations in class of a technical paper on some aspect of AI;
book or movie reports related to AI;
software review of software using AI techniques;
written surveys or

class presentations of several technical papers on AI;

original short stories which use AI in some essential manner.

Each student must choose two different types of larger assignments. The proposal for the first is due by September 20th, and the assignment itself is due by October 19th. In the case of an in-class presentation, the proposal should include a (pre-negotiated) date for the presentation, which will occur by October 19th.

The proposal for the second is due November 8th, and the final talk or demo must be done before classes end; papers, reviews, and stories may be submitted by the time of the scheduled exam (December 11th, 1pm).

Please make sure to look at the discussion of co-operation on projects and plagiarism.

Programming Projects

Programming projects should extend your understanding of the techniques discussed in class and your knowledge about the effectiveness of these techniques. They may be done in any language available to the instructor, for grading purposes. You will be graded on understandability of your work; you will write up specs, a justification of those specs, and a description of how you met them. Your program must run in order to receive credit.

In-Class Presentations

In-class presentations will use either transparencies or a computer (your choice of software). They will be rehearsed carefully and thoroughly in front of a live audience that gives feedback; quality of presentation will be graded as well as content. You should choose material of technical interest, demonstrate your understanding of the material, and leave your audience with an understanding of most of the material presented.

Sources of technical information

Your material should come from a reputable and refereed source, preferably a refereed journal or conference proceedings. Tech reports and random papers found on the web are not acceptable.

When a paper is submitted to a refereed journal, it is read carefully for correctness and presentation by 2--5 experts in the field. It may be read by these experts in several versions before being accepted for publication. It takes anywhere from 4 months to several years from initial submission to publication.
When a paper is submitted to a refereed conference, it is read for innovation, appropriateness to the conference, and sometimes for correctness by 1--3 experts in the field. These referees do not see the version that actually gets printed until it is published. It takes anywhere from 3 to 6 months from initial submission to publication.
The purpose of a tech report is to get a time-stamp on your work. They are not refereed. They "appear" within days of submission.
Anyone can post a paper on the web. Many do.

Reviews

Book or movie reports will cover a work that uses AI as an integral part of the plot. You will describe the plot and the role played by AI; analyse the author(s)' concept of AI, and conclude by discussing the feasibility of this concept. Discussions of social ramifications are welcome but not required. You will be graded on the thoughtful analysis of AI in this work; readability, prose structure, and technical details of writing (spelling, punctuation, etc.) will be a non-negligible part of the grade.

AI-Related Novels

Software Reviews

Software reviews can focus on one or several programs such as games that use AI techniques in some substantial manner. You will describe the overall software: What does it do, and how? You will describe the goal of the AI component, and the specific AI techniques used for that goal. Does the software succeed in its apparent or stated goals? Could those goals realistically be strengthened or extended, based on what you/we know about AI? What should researchers be working on the further extend the goals using not-yet-developed technology?

Surveys

A survey should cover three or more technical papers, preferably those published in refereed conferences or journals, on related topics. For instance, you could contrast three different constructions of neural nets to solve closely related problems, contrast several very different learning approaches to one problem, or compare the behaviour of one technique on several different problems. You must justify your choice of those particular papers; saying "they were Google's top choices" will not be an acceptable justification. Your survey should discuss the underlying problem (the application or technique that is the focus of your survey) and the choice of papers, then describe each of the papers, compare them, and come to some conclusions. You will be graded on the thoughtful analysis of AI in this work; readability, prose structure, and technical details of writing (spelling, punctuation, etc.) will be a non-negligible part of the grade.

See Sources

Short Stories

An original science-fiction story should have the usual features of a reasonable work of fiction: structure, plot, characterization, timing, and consistent point of view (or clearly defined shifts in that point of view). In addition, AI should play an integral role in the plot, without completely dominating the story. The type of AI should be identifiable from the story. It is up to the author whether or not the use of AI is realistic or fantastic, but even fantasy should be grounded in realistic technical detail. And as with all written work, the text should be clearly presented in 12-point type, left- and perhaps right-justified, without excessive use of fonts. There should be a title, author, and date on the first page in a larger font. Readability, prose structure, and technical details of writing (spelling, punctuation, etc.) will be a non-negligible part of the grade; the majority of points will be based on the plausibility, based on technical detail presented, of your imagined AI.

Cooperation

Group programming or survey projects are possible with consent of the instructor. Students may be encouraged to submit particularly good programming project write-ups or surveys for publication.

The usual applies: plagiarism, copying, or working together without crediting each other will be punished to the fullest extent of university regulations; homeworks due in class will not be accepted after class begins. If this is a problem, they may be submitted before class. Late homeworks accepted only with the usual documentation for illness or family crisis, or in the case of documented hardware failures.
Always back up your work, if possible to another machine or system.