A Knowledge Increasable Method for Artificial Neural Network

A neural learning system with the ability of knowledge increase is an important aspect of artificial neural network research,which is attracting the eyes of many machine learning researchers. Conventional research on knowledge increase/reuse of a neural learning system put an emphasis on the modification of a single network. The paper introduced the concept of autonomous artificial neural network （AANN for short） according to the idea of knowledge accumulation and succession. A neural learning system featuring knowledge increase ability is constructed with AANN modules,which is also flexible and scalable. An AANN module is different from its counterparts in that it is autonomous. With such modules as its basis,the reliability,flexibility and scalability of the system are greatly improved. Experimental results show that such a system can effectively inherit the knowledge of its components and can gain novel knowledge gradually when new modules arrive.