Exploration of Human Understandable Machine Learning in a Context Driven Modeling Architecture

摘要

The reliance on subject matter experts to provide expertise may restrict development of human behavior models for tactical tasks. Machine learning techniques have been used to forego or augment human expertise, however the knowledge typically generated by the machine learning algorithms is often not easily understandable by humans. The integration of two knowledge modeling methods, Context-based Reasoning and Contextual Graphs, provides an architecture that exhibits tactical behavior while representing human understandable knowledge. The application of a machine learning technique to develop the expertise within this architecture may result in behavioral models that maintain knowledge transparency.