Unsupervised Deep Learning with Autoencoders

Over the past few years, data-driven deep learning (DL) approaches have made impressive progress in the genomics field. The development of high-throughput technologies such as next-generation sequencing (NGS) has played a major part in this data-driven revolution. Several neural network (NN) architectures have found success in the genomics domain. For instance, in the previous chapters, we have seen feed-forward neural networks (FNNs), convolutional neural networks (CNNs), and recurrent neural networks (RNNs), which have been successfully used for many genomics applications. So far, all these NN architectures require that you have well-labeled data. However, a lack of ground truth and accurate labels is common in the genomics domain, which limits the application of supervised learning (SL) methods. NGS has significantly increased the use of gene expression assays, and there is so much genomics data out there with no label. Several methods...