THE REDUNDANCE OF MESSAGES IN THE NOGOOD LEARNING TECHNIQUES CASE

摘要

Asynchronous algorithms are characterized by the agents' use of the messages during the process of solution searching. The number of the recordings of nogood messages determines the exponential complexity of the algorithm in the unfavorable case in the first place. As a consequence of this fact, the calculus time and the asking for the hardware resources grow much faster than the dimension of the problem. Doubtlessly in the practical applications that use slow means of communication such as the Internet, the algorithms become inefficient because of the large number of the messages and their necessity for communications during the searching. This article analyzes the nogood learning techniques (combined with certain asynchronous techniques) from the prospective of the stream of redundant messages. There is a trial to identify the effects that the delays have in delivering messages on the performances of certain asynchronous techniques. In this article a filtering method for messages tails is proposed. This filtering method has a good effect on the message stream and the number of necessary cycles in order to obtain a solution. This filtering technique, applied for certain asynchronous techniques, allows an enhancement of the performances of asynchronous techniques.