Table 1.

Examples of DFNN network architectures tested in the random search. Hyperparameters shown include the regularization coefficient, number of layers, number of neurons per layer, and dropout fraction. Left column illustrates a simple network with fewer and smaller layers. Middle shows a complex network with more, larger layers. Right column shows the architecture of the highest performing network.

Simple dense network	Complex dense network	Highest performing dense network
L2 Regularization: 2.3e−4	L2 Regularization: 2.3e−4	L2 Regularization: 1.1e−4
Dense: 16 neurons	Dense: 128 neurons	Dense: 64 neurons
Dropout: 53% removed	Dropout: 18% removed	Dropout: 13% removed
Dense: 16 neurons	Dense: 128 neurons	Dense: 64 neurons
Decision Layer: 1 neuron	Dropout: 18% removed	Decision Layer: 1 neuron
	Dense: 64 neurons
	Dropout: 18% removed
	Dense: 42 neurons
	Decision Layer: 1 neuron