Sachiyo Arai

Learn More
Reasoning plays a central role in intelligent systems that operate in complex situations that involve time constraints. In this paper, we present the Adaptive Logic Interpreter, a reasoning system that acquires a controlled inference strategy adapted to the scenario at hand, using a variation on relational reinforcement learning. Employing this inference(More)
The point w e w ant t o m a k e in this paper is that Proot-sharingg a reinforcement learning approach i s v ery appropriate to realize the adap-tive behaviors in a multi-agent e n vironment. We discuss the eeectiveness of Proot-sharing theoretically and empirically within a Pursuit Game where there exist multiple preys and multiple hunters. In our context(More)
In this paper, we i n troduce FirstVisit Proot-Sharing (FVPS) as a credit assignment procedure , an important issue in classiier systems and reinforcement learning frameworks. FVPS reinforces eeective rules to make a n agent acquire stochastic policies that cause it to behave v ery robustly within uncertain domains, without pre-deened knowledge or subgoals.(More)
In this paper we describe Icarus, an architecture for physical agents that uses hierarchical skills to support reactive execution. We review an earlier version of the system, then present an extended framework that associates reward with stored concepts and utilizes a model-based approach to select among instantiated skills. Learning involves estimating(More)
In this paper, we discuss Proot-sharing, an experience-based reinforcement learning approach (which is similar to a Monte-Carlo based reinforcement learning method) that can be used to learn robust and eeective actions within uncertain, dynamic, multi-agent s y s t e m s. W e i n-troduce the cut-loop routine that discards looping behavior, and demonstrate(More)