Transcript Master`s Projects Overview

Topics for
Master’s Projects and Theses
-- Winter 2003 -Franz J. Kurfess
Computer Science Department
Cal Poly
Email: [email protected]
© 2000-2001 Franz Kurfess
Project Topics 1
Areas of Interest
 Artificial
Intelligence
 knowledge
management
 neural networks for structured knowledge
 Software
Engineering
 component-based
systems
 process-oriented design
 Educational
 teaching
Systems
support
© 2000-2001 Franz Kurfess
Project Topics 2
Specific Topics
 Knowlets
 Content-Based
Spam Filtering
 Ontologies
 Neural
Networks for Structured Knowledge
 How Computers Work
 AI Toolbox
© 2000-2001 Franz Kurfess
Project Topics 3
Knowlets
 component-based
systems for knowledge
management
 modular
design of more complex systems from simpler
components
 similar to components in Software Engineering, but the
emphasis is on content, not function
 content-based organization of knowlet collections
 Background: AI,
SE, possibly data bases
 Funding: possibly through CAD-RC
 other
funding sources under investigation
© 2000-2001 Franz Kurfess
Project Topics 4
Content-Based Spam Filtering
 design
and implementation of a system that categorizes email
messages

goal: identify as many unsolicited email messages (“spam”) as
possible

avoid false positives (valid messages mis-classified as spam)
 methods

use content-based and possibly usage-based techniques rather than
explicit rules to filter out spam

Bayesian networks


http://www.paulgraham.com/spam.html
Collaborative filtering
 Background:
AI, SE; possibly in combination with knowlets
 Funding: None
© 2000-2001 Franz Kurfess
Project Topics 5
Ontologies
 design
and implementation of ontologies
 (semi-)automatic
extension of ontologies
 user interfaces for ontologies

various perspectives, presentation methods
 Background: AI,
SE
 Funding: some through CAD-RC
 other
funding sources under investigation
© 2000-2001 Franz Kurfess
Project Topics 6
Neural Networks for Structured
Knowledge
 processing
of complex structures representing
knowledge with neural networks
 most
NNs are based on vectors, and can’t represent
knowledge easily
 recurrent NNs are more powerful, but also more difficult to
handle
 experimentation with various types of NNs to evaluate their
suitability


bio-informatics (drug discovery, genome sequencing)
knowledge management (ontologies, relationships between
documents)
 Background: AI,
specific domain
 Funding: None now
© 2000-2001 Franz Kurfess
Project Topics 7
How Computers Work
 demonstrations
and animations of important
concepts in computer science
 goal:
visualize and animate abstract concepts and
methods that may be difficult to understand from static text
and diagrams
 hardware

CPU, memory, hard disk, …
 OS

algorithms
CPU scheduling, disk scheduling, memory management, deadlock
detection, …
 data
structures and algorithms
 Background:
Java
 Funding: Would be most welcome :-)
© 2000-2001 Franz Kurfess
Project Topics 8
AI Toolbox
 generic
educational environment
 agents

perform tasks in a simulated or real environment
search for a goal, explore a room, perform a task, chase other
agents, work in teams, …
 development
of search algorithms, games, knowledge
representation methods
 modular design

playground, agents with various capabilities
 Background: AI,
 Funding:
SE
None
© 2000-2001 Franz Kurfess
Project Topics 9