Learning of recurrent convolutional neural networks with applications in pattern recognition

Convolutional neural network (CNN) has been successfully applied in character recognition. To further reduce the error rate of classification, based on traditional CNN, a recurrent-type CNN (RCNN) is presented in this paper. The Elman-Jordan recurrent model is embedded in the full connection layer of the proposed CNN. By optimizing the structure of the traditional CNN and making full use of the better learning ability of recurrent neural network, the proposed RCNN has a better recognition ability for input signals with noises. To verify the performance of the developed RCNN, some experiments are accomplished on the Chinese car plates and MNIST datasets. The experimental results show that, compared with traditional CNN and Elman-type CNN, a much smaller error rate can be guaranteed by our model.