• Summary
• Related Documents
  • Active Bibliography
  • Co-citation
• Version History

Abstract:

Abstract. We propose a genetic algorithm to accomplish belief revision. The algorithm implements a new evolutionary strategy resulting from a combination of Darwinian and Lamarckian approaches. Besides encompassing the Darwinian operators of selection, mutation and crossover, it comprises a Lamarckian operator that mutates the genes in a chromosome that code for the believed assumptions. These self mutations are performed as a consequence of the chromosome phenotype's experience obtained while solving a belief revision problem. They are directed by a belief revision procedure which relies on tracing the logical derivations leading to inconsistency of belief, so as to remove the latter's support on the gene coded assumptions, by mutating the genes. The algorithm, with and without the Lamarckian operator, has been tested on a number of belief revision problems, and results show that the addition of the Lamarckian operator improves the efficiency of the algorithm. Additionally, the combination of Darwinian and Lamarckian operators is useful not just for standard belief revision problems, but for problems where different chromosomes may be exposed to different constraints and observations. In these cases, the Lamarckian and Darwinian operators play different roles: the Lamarckian one is employed to bring a given chromosome closer to a solution (or even to find an exact one) to the current belief revision problem, whereas the Darwinian ones exert the role of randomly producing alternative chromosomes in order to deal with unencountered situations, by means of exchanging genes amongst them. We also tested this hypothesis, with positive results, for the case of a multi-agent belief revision setting. 1.

Citations