A simple and effective method for image classification

As a deep learning model, convolutional neural network (CNN) has greatly attracted attentions from researchers and has found its successful applications in many fields such as computer vision, pattern recognition, natural language processing, etc. However, the training of deep convolutional neural networks (DCNN) is very time-consuming and memory intensive. Inspired by the idea of extreme learning machine (ELM), this paper proposed a simple and effective method for image classification. The proposed method employs a neural network with 6 layers to classify images. The neural network consists of two modules: CNN and ELM network. The CNN has two convolutional layers and two pooling layers (also called subsampling layers). The ELM network is a single-hidden layer feed-forward neural network (SLFN). As does in ELM, the parameters of convolutional kernels of CNN are also randomly assigned. Our experimental results show that although this method is simple, it is very effective.