Open Access

Table C.1.

Hyperparameters of NEAT-VAE.

Hyperparameter Sampling sets
Layers {6, 8, 10, 16, 24}
Hidden neurons {30, …, 37}
Bottleneck neurons {10, …, 35}
μN { − 11, −9.21, −8, −5, −1}
σN Unif[1, 2]
LR network weights {0.005, 0.001, 0.0005, 0.01}
LR {0.0025, 0.0005, 0.00025, 0.005}
Batch size {16, 64, 128, 256, 512}

Notes. The number of layers, neurons per layer and bottleneck neurons determine the network architecture. μN and σN are transformation parameters of the noise covariance matrix N. The optimization of learnable parameters is determined by the learning rates (LR) for network weights ϕ, and the latent noise , which are tuned to minimize the objective function in Eq. (B.8). When using mini-batching, the batch size determines how many data samples are used to compute the loss function before back propagation and model updating is performed.

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.