A deep learning based technique for training deep convolution neural networks

摘要

Problem to be solved: to provide a method of constructing a convolution neural network based classifier for variant classification, a non temporary computer readable storage medium and a system.A convolution neural network based classifier for variant classification uses a backpropagation based gradient update technique that is progressively matched with ground truth labels to provide a group of residual blocks.Each group of residual blocks is parameterized by the number of convolutional filters in the residual block, the convolutional window size of the residual block, and the expansion convolution rate of the residual block.The size of the convolution window varies between groups of residual blocks, and the expansion convolution rate varies between groups of residual blocks.The training data includes benign training examples and virulence training examples of transforming sequence pairs generated from benign variants and virulent variants.Diagram