Coherent Knowledge Processing at Maximum Entropy by SPIRIT

TL;DR

SPIRIT is an expert system shell that manages probabilistic knowledge bases by generating maximum entropy distributions, supporting knowledge acquisition, rule formulation, and inductive learning for medium-sized applications.

Contribution

Introduces SPIRIT, a novel expert system shell that efficiently processes probabilistic knowledge with maximum entropy, including user-friendly tools and inductive learning capabilities.

Findings

Successfully handles medium-sized applications.

Supports inductive learning.

Generates maximum entropy probability distributions.

Abstract

SPIRIT is an expert system shell for probabilistic knowledge bases. Knowledge acquisition is performed by processing facts and rules on discrete variables in a rich syntax. The shell generates a probability distribution which respects all acquired facts and rules and which maximizes entropy. The user-friendly devices of SPIRIT to define variables, formulate rules and create the knowledge base are revealed in detail. Inductive learning is possible. Medium sized applications show the power of the system.