| 摘要 |
A computer system comprising a computer having a calculation device and a memory, and also comprising: a graphical model construction unit that constructs a graphical model by using learned data; and an inference unit that receives input of evidence data and calculates an inference result whereby an inference target node has a probability value distribution having a prescribed state value. The inference unit includes: an impact calculation unit that calculates impact evaluation values; and a probability value derivative amount calculation unit that calculates probability value derivative amounts. The probability value derivative amount calculation unit: extracts an inference target node and dependent nodes; selects a node to be processed, from among the extracted nodes; obtains a probability table for the processing-target node; extracts, as a family node, a set including the processing-target node and at least one dependent node; calculates a simultaneous probability distribution for a simultaneous probability variable combining the inference target node, the processing-target node, and the family node; and calculates a probability value derivative amount using the simultaneous probability distribution and the processing-target node probability table. |
