The Neural-Network-Based Student Simulator ("STU") is designed
to provide a variety of learning capabilities to the adaptive
tutor.
The Neural-Network-Based Student Simulator ("STU") is designed
to provide a variety of learning capabilities to the adaptive
tutor.
Here is a functional schematic of STU:
Each component of the student simulator serves as a model of
one of the functional elements involved in learning.
The Perception Processor accepts input from the tutor and changes
it into the student's internal representation. A probability matrix
specifies the "odds" of how a given input digit will be
represented (or misrepresented) in the student's mind.
The perception accuracy can be varied using a single parameter.
The Central Processor controls the interactions among all the
student simulator modules, and performs simple symbolic tasks
such as copying and comparing symbols or accessing memory.
The Working Memory module models the "Short Term Memory":
it can learn quickly, but the memories also fade quickly.
A special symbol-oriented "Working Memory Network" has been written
to perform this function.
The Long-Term Memory (LTM) models the human
student's "permanent storage". This module uses the CNLS
Network which "learns" by adjusting "weights". The LTM access can be optionally
set to two modes: access through the Working Memory only, or direct access from
the Central Processor.
The Response Processor is the inverse of the Perception Processor. It
translates the answer from the student's mind into the form of a typed
response to the tutor, again using a probability matrix that sets the
odds of obtaining a correct or incorrect translation. In addition, the
Response Processor has a model for the response time that is based on a
combination of the student's knowledge certainty and a mechanical typing
time that is derived from actual keyboarding measurements.
The student simulator
is missing some very important aspects of human capabilities. Perhaps the most
significant omission is the lack of any kind of "metacognition" or mechanism for
self understanding and high-level adaptation. This is a fascinating and important
area for future research.
Return to Phase I Results Page or ANSR / DOEd SBIR Project Page
Return to ANSR Home Page or Site Contents -- or -- Contact ANSR.