Hybrid Models for Sequential Decision Making.

Abstract

To study hybrid architectures of complex learning, especially as applied to a simulated minefield navigation task, we developed a hybrid connectionist model CLARION as a demonstration of the approach of bottom-up skill learning (from procedural to declarative knowledge). The model essentially consisted of two levels for capturing both procedural and declarative knowledge and enabling bottom-up learning, which differed markedly from existing models. Our Computational and psychological experiments during the past five years demonstrated matches of the model with human data in several domains.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 15, 2000
Accession Number
ADA373424

Entities

People

  • Ron Sun

Organizations

  • University of Missouri

Tags

Communities of Interest

  • Autonomy

DTIC Thesaurus Topics

  • Artificial Intelligence
  • Artificial Intelligence Software
  • Behavioral Sciences
  • Cognitive Science
  • Consciousness
  • Engineering
  • Information Processing
  • Information Science
  • Intelligent Agents
  • Intelligent Systems
  • Learning
  • Machine Learning
  • Multiagent Systems
  • Navigation
  • Neural Networks
  • Psychology
  • Reinforcement Learning

Fields of Study

  • Computer science

Readers

  • Artificial Intelligence