Knowledge-increasable model and its algorithm
Yingjian Qi, Jian-Yu Li, Zhao Jiali, Luo Siwei · 2002
A knowledge-increasable system (KIS) is a kind of combined artificial neural network. This conception is proposed to accomplish inheritance and accumulation of the sub-network without destroying the existing function. We put up a model to perform the KIS, and compare this model with the popular Mixture of Experts (ME) model. We classify the model by three different situations and give the corresponding algorithm. We introduce future work.