Fully Kernected Neural Networks

In this paper, we apply kernel methods to deep convolutional neural network (DCNN) improve its nonlinear ability. DCNNs have achieved significant improvement in many computer vision tasks. For an image classification task, the accuracy comes saturation when depth and width of are enough appropriate. The will not rise even by increasing width. We find that improving ability can break through accuracy. a DCNN, former layer is more inclined extract features latter classify features. Therefore, at last fully connected implicitly map higher-dimensional space so achieves better linear separability. Also, name as kernected networks (fully with methods). Our experiment result shows achieve higher faster convergence rate than baseline networks.