Fig. 1: An overview of machine learning scenarios and commonly used DL architectures.
From: Current progress and open challenges for applying deep learning across the biosciences
Top panel encapsulates the three most common paradigms of machine learning: supervised learning in which dataset contains ground truth labels, unsupervised learning in which dataset does not contain ground truth labels, and reinforcement learning in which an algorithmic agent interacts with a real or simulated environment. The bottom panels provide an overview of the most prevalent DL architecture ideas each designed to achieve specific highlighted goals. An additional set of short descriptions is provided for other common components of DL architectures mentioned in the manuscript.
