Probably Approximately Correct Heuristic Search
نویسندگان
چکیده
A* is a best-first search algorithm that returns an optimal solution. w-admissible algorithms guarantee that the returned solution is no larger than w times the optimal solution. In this paper we introduce a generalization of the w-admissibility concept that we call PAC search, which is inspired by the PAC learning framework in Machine Learning. The task of a PAC search algorithm is to find a solution that is w-admissible with high probability. In this paper we formally define PAC search, and present a framework for PAC search algorithms that can work on top of any search algorithm that produces a sequence of solutions. Experimental results on the 15-puzzle demonstrate that our framework activated on top of Anytime Weighted A* (AWA*) expands significantly less nodes than regular AWA* while returning solutions that have almost the
منابع مشابه
Search-Aware Conditions for Probably Approximately Correct Heuristic Search
The notion of finding a solution that is approximately optimal with high probability was recently introduced to the field of heuristic search, formalized as Probably Approximately Correct Heuristic Search, or PAC search in short. A big challenge when constructing a PAC search algorithm is to identify when a given solution achieves the desired sub-optimality with the required confidence, allowin...
متن کاملImproving Query Correctness Using Centralized Probably Approximately Correct (PAC) Search
A non-deterministic architecture for information retrieval, known as probably approximately correct (PAC) search, has recently been proposed. However, for equivalent storage and computational resources, the performance of PAC is only 63% of a deterministic system. We propose a modification to the PAC architecture, introducing a centralized query coordination node. To respond to a query, random ...
متن کاملOn Efficient Heuristic Ranking of Hypotheses
This paper considers the problem of learning the ranking of a set of alternatives based upon incomplete information (e.g., a limited number of observations). We describe two algorithms for hypothesis ranking and their application for probably approximately correct (PAC) and expected loss (EL) learning criteria. Empirical results are provided to demonstrate the effectiveness of these ranking pro...
متن کاملA probabilistic approach to case-based inference
The central problem in case based reasoning (CBR) is to infer a solution for a new problem-instance by using a collection of existing problem-solution cases. The basic heuristic guiding CBR is the hypothesis that similar problems have similar solutions. Recently, some attempts at formalizing CBR in a theoretical framework have been made, including work by Hüllermeier who established a link betw...
متن کاملProbably Approximately Correct Search
We consider the problem of searching a document collection using a set of independent computers. That is, the computers do not cooperate with one another either (i) to acquire their local index of documents or (ii) during the retrieval of a document. During the acquisition phase, each computer is assumed to randomly sample a subset of the entire collection. During retrieval, the query is issued...
متن کامل