A Covering Algorithm Based on Competition
Qingsheng Wang, Yueqin Zhang, Xianyang Sun, Bin Liang · Procedia Engineering · 2012
In covering algorithm, the order of learning samples is random, but experiments show that the learning sequence has significant impact on the network performance. This paper proposes a new algorithm which is covering algorithm based on competition (CAC). In this algorithm, sphere neighborhoods can be adjusted gradually, the ill-suited sphere neighborhoods will be removed and the whole neural network will performs more stable. Finally the algorithm is applied to a widely used database. The experiment results show that it can effectively decrease the number of rejected samples, reduce the number of sphere domains and improve the recognition accuracy.