A novel hybrid learning scheme for pattern recognition

This paper presents a novel hybrid learning scheme applicable to pattern classification tasks. The proposed self-supervised learning method is a combination of unsupervised clustering k-means fast learning artificial neural network (KFLANN) and a typical supervised backpropagation learning algorithm (BP Network). Complementary benefits of both algorithms motivated the development of the hybrid model. The KFLANN clustering output is used as the target value for BP training, resulting in an efficient self-supervised learning model. Experimental results are comparable with non-hybrid individual algorithm.