get_possible_arch

behavenet.models.ae_model_architecture_generator.get_possible_arch(input_dim, n_ae_latents, arch_seed=0)

Generate a random autoencoder architecture.

Parameters:

• input_dim (array-like) – dimensions of batch with shape (n_channels, y_pix, x_pix)

• n_ae_latents (int) – number of autoencoder latents

• arch_seed (int, optional) – set rng seed to reproduce architecture generation

Returns:

dictionary that fully defines autoencoder architecture, which is used as input to the AE model in behavenet.models.aes

Return type:

dict