Collecting Active Information from Intelligent Agents
Collecting Active Information from Intelligent Agents to Improve Search Effectiveness
Eric Holloway (Air Force)
Dembski's "Search for a Search" implies that intelligent agents can introduce information into the search process (1). This means that intelligent agents may not be bound by the No Free Lunch Theorem. If true, then incorporating the input of intelligent agents into a search algorithm should improve the results. My study benchmarks the performance of search and optimization algorithms against algorithms that also have human input during the search process. The criteria of comparison is the amount of information each process has about the search domain, and which process finds better solutions based on this information. The amount of information each process has is measured by the rate of solution improvement per solution evaluation.
(1) Dembski, W. A. 2005. Searching Large Spaces: Displacement and the No Free Lunch Regress.