*************************************************************
Autonomous Learning Agents
*************************************************************
Time Table
------------
Introduction and Definitions: 2 weeks
Model abstraction ("learning" and "discovery"): 6 weeks
Model application ("doing"): 3 weeks
Integration of "learning" and "discovery" with "doing": 2 weeks
System analysis: 2 weeks
Course project: 1 week
Main References
---------------
Autonomous Learning from the Environment, Computer Science Press, 1994,
plus various papers on each of the techniques discussed in the class.
Lecture Plan
--------------
1. Jan 10, 1996. Course organization, philosophy, history, background.
Definitions of Autonomous learning: actions, percepts, mental
constructors, models, transparent and translucent environments, etc. Ch.
1,2 and related references.
2. Jan 17, 1996. Tasks of autonomous learning, and views of other
scientific fields, such as function approximation, adaptive control,
cognitive psychology, and others. Ch. 3 and related references.
3. Jan 24, 1996. Model abstraction in transparent environment, Direct
recording, the problem and algorithms for abstracting from
attribute-based perceptions, and active learning of concepts. Ch.
4.1-4.5, 4.7-4.9, and
related references.
4. Jan 31, 1996. Model abstraction in transparent environment, the
problem and algorithms for abstracting from structure or relation-based
perceptions. Ch. 4.6, 4.10, and related references.
5. Feb 7, 1996. Model abstraction in transparent environment, Bayesian
Probability, neural networks and other uncertainty reasoning techniques.
Ch. 4.11-12 and related references.
6. Feb 14, 1996. Model abstraction in translucent environment, Problems
and active learning finite automata with reset, the L* and related
algorithm. Ch. 5.1-5.2 and related references.
7. Feb 21, 1996. Model abstraction in translucent environment, active
learning finite automata without reset, homing sequences and local
distinguishing experiments. Ch. 5.3-5.8 and related references.
8. Feb 28, 1996. Model abstraction in translucent environment,
Stochastic automata and Hidden Markov models. Ch. 5.9-5.10 and related
references.
9. Mar 6, 1996. Model application, searching for optimal solutions,
Dynamic programming, A*, Q-learning. Ch 6.1 and related references.
10. Mar 13, 1996. Model application, searching for satisficing
solutions, Real-time A*, Distal supervised learning, and symbolic goal
regression. Ch 6.2 and related references.
11. Mar 20, 1996. Model application, designing and learning from
experimentation. Ch 6.3-6.5 and related references.
12. Mar 27. Integration of model abstraction and model application in
transparent environment. Ch. 7.1.1, 7.2.1, and related references.
13. Apr 10. Integration of model abstraction and model application in
translucent environment. Ch. 7.1.2, 7.2.2, and related references.
14. Apr 17. Systems. LIVE architecture and details. Ch. 8, 9, 10, and
related references.
15. Apr 24. Systems. LIVE performance (continues), and systems for Map
Learning. Ch. 11, 12, and related references.
16. May 1. The future of autonomous learning systems, and project
discussion (project assignment will be given on Mar 27).