Study of full interval feed-forward neural network
Shouping Guan, Rong-ye Liang · 2016
This paper proposes an architecture and BP learning algorithm for interval feed-forward neural networks (IFNNs)with all parameters setting to intervals, which can be called full interval feed-forward neural network(FIFNN), to handle uncertain information represented in interval-valued formation. In order to evaluate the performance of FIFNN, we chose two testing functions with different complexity to be modeled, and compare with other types of IFNNs. The simulation results show that the converging and the approximating ability of our proposed network are better than the other IFNNs' models in some situations.