Figure 2.

An example of the deep adversarial autoencoder structure. The deep adversarial autoencoder structure comprises two main components including an autoencoder module and an adversarial network module. The autoencoder module comprises an encoder unit and a decoder unit. The encoder unit also serves as the generative network module of the adversarial network architecture. Step 1: The encoder unit produces synthetic instances as real as possible. One particular example in drug design and discovery is a reconstructed latent vector as a fake instance. Step 2: The discriminative network module assesses the probability that an instance stems from the real dataset. One particular example in drug design and discovery is a real latent vector from the drug-like compound dataset. Step 3: Both the autoencoder and discriminative network modules play concurrently against each other to obtain their objectives.