Semisupervised Learning to Train Ensembles of Deep Convolutional Neural Networks

摘要

The disclosed technology relates to the construction of a convolutional neural network-based classifier for variant classification. Specifically, this technique relates to training a convolutional neural network-based classifier on training data using a backpropagation-based gradient update technique that incrementally matches the output of a convolutional neural network-based classifier with a corresponding ground truth marker. it''s about A convolutional neural network-based classifier comprises a group of residual blocks, each group of residual blocks parameterized by the number of convolution filters in the residual block, the convolution window size of the residual block, and the atros convolution rate of the residual block. , the size of the convolution window varies between groups of residual blocks, and the atros convolution rate varies between groups of residual blocks. The training data includes positive training examples and pathogenic training examples of translated sequence pairs generated from positive and pathogenic variants.