Particle swarm optimization for deep learning of convolution neural network

A Deep-learning architecture is a representation learning method with multiple levels of abstraction. It finds out complex structure of nonlinear processing layer in large datasets for pattern recognition. From the earliest uses of deep learning, Convolution Neural Network (CNN) can be trained by simple mathematical method based gradient descent. One of the most promising improvement of CNN is the integration of intelligent heuristic algorithms for learning optimization. In this paper, we use the seven layer CNN, named ConvNet, for handwriting digit classification. The Particle Swarm Optimization algorithm (PSO) is adapted to evolve the internal parameters of processing layers.