Illustration of the network architecture of the simple NN classifier (top row only, blue layers) and the DANN (full figure). Through the cooperation of both the encoder and the discriminator network, the DANN learns domain-invariant features in the embeddings layer (light blue). The DANN is trained adversarially to deceive its discriminator into being unable to distinguish source from target domain instances. The figure also shows the dimensionality of the input (number of instances, N_i, by 5) and the number of neurons in each layer (in bracket next to each layer name).