Learning with Regularization Networks Mixtures

Abstrakt

We propose a supervised learning model – regularization networks mixtures. It consist of unsupervised component realizing partitioning of the input space and a mixture of regularization networks. We show that the model has much lower time requirements with generalization ability comparable to a single regularization network.

Kategorie a štítky

  • Konferenční příspěvek (mezinárodní konf.)