Summary

The application of unsupervised DL through learned representation is becoming extremely popular in genomics because of the large-scale datasets produced using NGS technologies. Autoencoders are being routinely used by researchers because of their promise and success across many genomics applications. Autoencoders learn by the reduced representation of the data through compression and reconstruction. During the process, they learn the key features of the data and identify the data structure automatically from examples rather than through handcrafting by humans. Diverse types of autoencoders exist to ensure that the reduced representation of the data identifies the key attributes of the original data. Autoencoders have several applications in genomics, mainly in gene expression analysis. With tools such as ADAGE, autoencoders are helping genomics datasets with no labels get biological insights from that data. We started the chapter by understanding what is unsupervised learning...