ncem.models.ModelEDncem

class ncem.models.ModelEDncem(input_shapes, latent_dim=10, dropout_rate=0.1, l2_coef=0.0, l1_coef=0.0, enc_intermediate_dim=128, enc_depth=2, dec_intermediate_dim=128, dec_depth=2, cond_type='gcn', cond_depth=1, cond_dim=8, cond_dropout_rate=0.1, cond_activation='relu', cond_l2_reg=0.0, cond_use_bias=True, use_domain=False, use_type_cond=False, scale_node_size=False, output_layer='gaussian', **kwargs)[source]

Model class for NCEM encoder-decoder with graph layer IND (MAX) or GCN.

Methods

Parameters

  • latent_dim (int) –

  • dropout_rate (float) –

  • l2_coef (float) –

  • l1_coef (float) –

  • enc_intermediate_dim (int) –

  • enc_depth (int) –

  • dec_intermediate_dim (int) –

  • dec_depth (int) –

  • cond_type (str) –

  • cond_depth (int) –

  • cond_dim (int) –

  • cond_dropout_rate (float) –

  • cond_activation (Union[str, Layer]) –

  • cond_l2_reg (float) –

  • cond_use_bias (bool) –

  • use_domain (bool) –

  • use_type_cond (bool) –

  • scale_node_size (bool) –

  • output_layer (str) –