| 摘要 |
PROBLEM TO BE SOLVED: To realize parameter estimation at a higher speed than that at the time of operating the processing in a batch from all data at the time of estimating the parameter of a probability model. SOLUTION: This system is provided with a synthetic learning mechanism 11 and plural agents 1-(s). The agents 1-(s) which are connected with distributed terminals estimate the parameter of a probability model from input data according to Bayesian posterior probability. The synthetic leaning function 11 inputs the estimated values of all the agents 1-(s), and estimates the parameter of the probability model. Then, the estimated result of the synthetic learning function 11 is feedbacked to each agent 1-(s) and the estimating process is repeated. Thus, the parameter estimation can be operated in parallel by the plural distributed terminals, and the parameter estimation can be attained at a high speed by synthesizing this result. |
