Research In Multi-Agent Learning Algorithms

广泛的理论问题处理多智能体学习系统将被调查。当涉及多个代理时，合作是成功完成任务的关键。注意力将局限于多个智能体合作学习一个概念（或功能）的学习系统。最终目的是设计出合作学习策略的模板，希望能应用于其他实际问题。该目标将通过研究具有不同收敛性和/或正确性标准的学习器以及其他特征（如假设程序的总体和资源有界性）来实现。这种广泛的学习器类型将有助于识别人们在多智能体学习系统中应该期望的合作的真实性质和复杂性。特别是，对fin类型（以及随后的其他类型）学习中的概率和多元层次结构的持续研究将继续进行，该研究最近揭示了一种新的但非常密切的合作策略。

A broad range of theoretical issues dealing with multi-agent learning systems will be investigated. Cooperation is the key to successfully completing a task when more than one agent is involved. Attention will be restricted to learning system where multiple agents cooperate in learning a concept (or function). The ultimate goal is to devise templates for cooperating learning strategies which hopefully can be applied to other practical problems. The goal will be approached by investigating learners with differing convergence and/or correctness criteria as well as other features such as totality of hypothesized programs and resource boundedness. This wide spectrum of type of learners will help in identifying the true nature and complexity of the cooperation one should expect in multi-agent learning systems. In particular, an on-going investigation of the structure of probabilistic and pluralistic hierarchy in FIN-type (and subsequently in other types of)learning will be continued, which recently revealed a new but very intimate cooperative strategy.