abstract |
A pattern learning model having scale extensibility and generalization characteristics can be obtained. A learning module 10 1 to 10 N, using the input data to perform update learning to update a plurality of model parameters of a pattern learning model. The model parameter sharing unit 20 to the learning module 10 1 to 10 N, to share a model parameter. Shared intensity control unit 112, for each learning module 10 1 to 10 N, the shared intensity beta ij from the learning module 10 j represents the intensity of the impact of the learning module 10 i, and updates a plurality of model parameters by updating the learning Sometimes the learning module 10 i is updated to minimize the learning error E i . The present invention can be applied to, for example, time-series pattern learning. [Selection] Figure 12 |