A probabilistic inference method with multiple evidences and its implementation using a layered network

The inference method can deal with multiple ambiguous evidences, and can describe effectiveness of evidences and relationships between evidences. Therefore, it is more advantageous than the Dempster-Shafer theory because of its ability to describe relationships between evidences. This method divides the whole world into possible worlds according to element value combinations. The probability of each possible world is decided so as to satisfy constraints corresponding to a priori knowledge and to maximize the entropy of the whole world. Furthermore, the method can be implemented using a layered network. In this network, individual network units do not have to perform complicated operations and connections between the layers are restricted. In other words, this network consists of simple units and restricted connections, thus, high speed processing will be possible using parallel processing