BUSS 925: Techniques for Knowledge-Based Systems Development
Department of Business
Systems, University of Wollongong
Lecture content outlines and pointers to web resources:
- Lecture 1: We looked at definitions of AI/KBS's, the history of the
area, major milestones and outlines of the major sub-areas. The Association for Computing Machinery
(ACM) Special Interest Group on Artificial Intelligence (SIGART)
directory of web resources on AI is a great place to start
exploring. The ACM is the oldest and
largest association of computing professionals in the world, and its
homepage is worth looking at. The SIGART
homepage contains other interesting information, including a pointer
to the SIGART Bulletin which contains good survey articles, as well
announcements of new conferences/seminars.
- Lecture 2: We considered a generic view of AI as the design and
construction of intelligent agents (for more information on research and
development in intelligent agents, look at the UMBC AgentWeb. We looked at the
KBS lifecycle and considered parallels
with standard software engineering and the systems development
lifecycle. This page contains
a good collection of pointers to web resources on software engineering.
- Lecture 3: We discussed the architecture of the MYCIN expert system
and looked
at a transcript of a session with MYCIN. You can download a variety of
interesting AI software (including a version of the EMYCIN system -
precursor to VPExpert - that we discussed in class and the ELIZA system
which simulates a psychotherapy session - also discussed in class) from
the
Carnegie Mellon University AI Repository. We talked about the Turing
Test for AI, and as promised, this is a pointer to a fun
contest where programs compete to win the Turing Test for true AI.
- Lecture 4: We discussed the EMYCIN and TEIRESIAS systems that
evolved from the MYCIN project. For more information on these other
related systems developed at Stanford University, click here.
- Lecture 5: We looked at reasoning with propositional logic.
You can try
out all of what we have done with propositional logic using this
interactive web-based propositional theorem prover developed by the
Information Systems group at the University of Newcastle - the system
is called VADER and you can access it by clicking here.
We also looked at various methods of representing uncertainty.
This as well
as this are
interesting pointers to the use of Bayesian reasoning techniques in
building Answer Wizards for Microsoft products.
- Lecture 6: We looked at reasoning with predicate logic (also
referred to as first-order logic). Although this might be a bit
complicated, check out this
webpage containing code for a predicate
logic theorem prover.
- Lecture 7: We looked at the general notion of search through a
state space as a common theme underlying most AI problem solving. We
considered blind search technqiues such as depth-first (DFS),
breadth-first
(BFS) and iterative deepening search (this link provides a demo of what
happens when depth-first search is applied to the problem of solving the
8-puzzle, while this link
provides applet-based demos of both DFS and BFS). We also looked at
heuristic se
arch techniques such as
best-first search and A* search (this link applies A* search to the 8-puzzle).
- Lecture 8: We looked at search techniques developed for solving
optimization problems, including hill-climbing and simulated annealing.
We also looked at how 2-person adversarial games are played using the
minimax algorithm. This is a
link to the Deep Blue system and its historic match with world chess
champion Gary Kasparov. This
is a link to the World (Man-Machine) Checkers Champion, a program called
Chinook.
- Lecture 9: We looked at the notion of intelligent agents. The NASA Remote Agent Experiment is very
interesting and worth following.This is a link to our (Dept.
of Business Systems Decision Systems
Lab) soccer team (called Gongeroos), consisting of 11
intelligent soccer-playing agents, which came fifth in Robocup'98 and
has just qualified (as one of only two Australian teams) for Robocup'99 (the
qualifying round had a culling rate of 50%). The UMBC AgentWeb is great source
of information on intelligent agent-based systems.
You can try an
agent-based
marketplace, an
d an online automated auction using the
AuctionBot.
- Lecture 10: We looked at how agents plan. We specifically considered
the STRIPS approach to planning. This
page contains pointers to a wide variety of AI planning resources.
- Final lectures: We looked at more issues related to agent-mediated
electronic commerce. We also learnt about constraint programming. This
is an easy-to-understand online guid to constraint programming. The
Constraints Archive site is worth exploring and this
commercial site
has a very useful collection of papers on applications of constraint
technology.
|