Gradual Information Acquisition in Multi-layered Self-adaptive Networks

R. Kamimura (Japan)


: mutual information maximization, gradual information acquisition, self-adaptive networks


In this paper, we propose a computational method called gradual information acquisition method to improve clas sification performance of self-adaptive multi-layered net works. In the gradual information acquisition method, multi-layered networks gradually obtain necessary infor mation in a course of learning. Thus, networks can dis cover different features corresponding to different levels of information. We applied the new method to the Iris data problem. Experimental results confirmed that information can gradually be maximized in multi-layered networks and that the classification performance is greatly improved.

Important Links:

Go Back