AI Seminar: Michael van Lent

[Eric D. Klavins]

AI Seminar
----------
Tuesday, Feb. 1, 4-5 PM
175 ATL (Large Conference Room)
Refreshments provided

Learning Task Performance Knowledge by Observation
--------------------------------------------------

Michael van Lent,
Graduate Student, AI Laboratory

Abstract:

Developing automated agents that intelligently perform complex real
world tasks is time consuming and expensive. The most costly part of
developing these agents is extracting knowledge from human experts and
encoding it in a form usable by automated agents. Automated learning
from a sufficiently rich and focused knowledge source can
significantly reduce the cost of development by automating the
knowledge acquisition and encoding process. Learning by observation is
particularly well suited to learning hierarchical performance
knowledge for tasks that require realistic, human-like behavior. Our
learning by observation system, called KnoMic (Knowledge Mimic), uses
a supervised learning algorithm that reads observation traces of an
expert performing a task and generates production rules that an agent
can use to perform the same task. Additionally, KnoMic detects many
types of errors in the learned knowledge and some can be corrected
through an instruction based interaction with the expert. Results of
our efforts to learn performance knowledge with KnoMic in the tactical
air combat domain are reported.