Universal Weak Methods in Soar
Universal Weak Methods in Soar
Soar was used to posit the existence of a universal weak method, an
approach that said that the search strategy should arise out of the
interaction between the structure of the agent and the proposed task.
The search strategy chosen is assumed to be weak; i.e., the
agent has little knowledge about the task environment. Thus, any of
the weak methods may arise in Soar out of the interaction of this
universal weak method and the task. The advantage of such an approach
is that it avoids program synthesis: the behavior results from the
interaction of knowledge and task rather than being explicitly
programmed. The following is a summary of some of the weak methods
that have been demonstrated in Soar using the idea of a universal weak
method (adapted from Laird, Newell and
Rosenbloom, (1987)):
- Heuristic Search
- Operator Subgoaling
- Waltz Constraint Propagation
- Means-End Analysis
- Generate and Test
- Breadth-First Search
- Depth-First Search
- Lookahead Search
- Simple and Steepest Ascent Hill-Climbing
- Progressive Deepening
- Mini-Max
- Alpha-Beta Pruning
- Iterative Deepening
- Branch and Bound
- Best-First Search
- Macro-operators
Return to the top of this architecture.
Go to a discussion of this capability
for multiple architectures.
Current Location: Soar-Capabilities-Universal Weak Methods