CS513 Applied Knowledge Engineering Professor Loui loui--aat--cs.wustl.edu TA's: none Location: Cupples-II 202 Time: Thurdays 5:30-7:30 (8:30 as needed) Course Description: see catalog Credit: 3.0 units Prerequisites: Senior-level programming skill in some platform, AI course (CS313. CS511, CS527, etc.) or equivalent. Required Texts: none Course in Brief: This course is not related to CS313, which is a scripted course with specific intellectual objectives. This is an applied project course, essentially a group Master's Project in AI. The main objectives of this class are practical and two-fold. First, students are to develop software and acquire data so that a real-world problem can be solved with real-world impact on real-world people. We try to define projects that have AI challenges or would benefit from AI approaches. But in the end, the grade depends on what the software does for whom, not what ideas it is based on or how it is written. Second, students report in the past that they benefit from sharing work styles, especially when students are part-time Master's students employed in industry, or are full-time students who have had research or industrial experience on large projects. We facilitate that kind of learning in this course. You may work with anyone in this class (within team-size parameters), use any data (within practical reason), use any processors or platforms, program in any languages, and use any software you can find that is helpful. The first three to four weeks of this course are dedicated to brainstorming, team-composition, identification of challenges, access to data and practical advice or expertise, and defining objectives for the software design. The middle of the semester is devoted to initial prototyping, review of the relevant literature, and testing. The final part of the semester is devoted to technology transfer and useability. Of all my courses, this course has received the highest average student satisfaction. Why? Requirements: Each student must participate in the proposal of projects. Some will be asked/allowed to present a proposal formally. Those who present must inquire about access to expertise and data, and become the de facto team leaders. You will have a formal requirement of identifying potential market competitors for software you might develop (1 page list). Each student must join a team within the first three weeks. Students may move from a team to another, and teams may be merged or split, but the fourth week is very late to be undecided about projects and team composition. Membership of the teams is decided by the professor. Each student who is not a team leader must either make a formal presentation of a proposed approach in the first three weeks, or a formal presentation of relevant literature, in the middle weeks of the semester. Teams will have a formal requirement of identifying potential patents or copyrights of relevance (1 page list). Each team must deliver a demonstration of the current state of their code (or data), each week, even if there is no substantial progress. (These start as soon as teams are launched.) Teams will have a formal requirement of presenting at l paper on a general method that is relevant to your project (selection to be discussed with the professor in advance). Your final project demonstration should include a persistent web site which can be visited after the semester ends (even if there are proprietary data or other IP considerations, in which case an abstract suffices.) Past Projects: digital image lint detector and repair, dead pixel removal, and GUI frame-to-frame semi-auto identification of aorta in video sequence the monkey left behind: sysadmin scripting for windows virus removal ------------- assistant for orthodontial x-ray geometric anomaly detection automatic satellite image tree-labeler ebay search expansion tool intelligent presentation of UNIX man pages real-time video alarm for monitoring solo night -shift staff engineers' assistant for interpreting pilot lingo during flight simulations expert system for automatic repair of payment transmission errors automatic detection of morphological features in 3d brain scans ------------- simulation of local traffic system for testing road sensor and light-switching algorithms automatic ranking of robot event photographs genetic algorithms for quarterly parts-inventory prediction in-flight negotiation simulator for UCAV's with component failures and mixed objectives ------------- automatic detection of species reporting errors in BLAST database neural nets for detection of specific features in MRI data simulated annealing for allocation of frequencies to metro phone cells My Suggestion(s): O(nlogn) clustering of documents (for stream processing) O(hw) detection of similar images parallel language corpus for MT and NL/KDD experiments foreign language document classification with english training lint removal real-time determination of focus and exposure based on high level semantic features (future digital camera interface) Your Suggestions: 17 18 19 ^today demo 20 21 22 23 24 25 26 ^no class 27 28 29 30 1 2 3 ^last official demo December 2005 S M Tu W Th F S 4 5 6 7 8 9 10 ^i am at JURIX 11 12 13 14 15 16 17 XXXX funding review ^last project review meetings th 5:30-8 (save a b, raise an a-) 18 19 20 21 22 23 24