Description of the Architecture

The graph depicts the structure of MAX. MAX is constructed of three layers: (1) the knowledge representation layer, (2) the architecture layer and (3) the knowledge layer. The first two domain-independent layers define how MAX learns and utilizes knowledge to attain goals. The knowledge layer defines the domain knowledge including facts and operators needed to manipulate facts and to learn.

Representation of Knowledge

MAX specifies that knowledge be composed of logic frames(l-frames), declarative structures comprising logical conjunctions of literals. These logic frames contain information regarding context-relevance, constraints, goal-relevance, preconditions, post-conditions, and the action to be performed in the form of a production. Nested l-frames can be utilized to explicitly encode meta-reasoning.

Architecture

The operating kernel of MAX, a procedural production system interpreter, uses many small production memories, arranged as procedures and called behaviors, to implement reasoning, learning or action. A rule-based forward chaining engine, this architecture allows only a single behavior to be active at a time. As shown in the diagram, these production memories are broken into the following categories:

Each region consists of sets of rules and operators. Rules specify which operator shall be used, while operators determine the adds and deletes to be performed in the current state and in the long term memory.

Knowledge Layer

This layer reputedly contains the necessary components for reasoning but its structure and use cannot be elicited from the description provided in the reference.

Return to the Top of this architecture.