| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide an inference device capable of obtaining a concrete inference value in the case where there is a node for which a continuous random variable is discretized and is determined as a state, when executing inference using the node as an inference node. <P>SOLUTION: An inference device 100 uses a probability model of a Bayesian network to infer a non-observed node that represents an event, for which a continuous value is observed, by a random variable in a discrete state from an observation node which has obtained an observed value. The device 100 comprises: an input section 40 for inputting the observed value of the observation node; a posterior probability calculation section 31 for obtaining posterior possibility distribution of the non-observed node, based on the observed value of the observation node input by the input section 40; a restoration section 32 for obtaining an inference value of the non-observed node, based on the posterior probability distribution; and an output section 50 for outputting the inference value obtained at the restoration section 32. <P>COPYRIGHT: (C)2013,JPO&INPIT |
