ncem.estimators.EstimatorEDncem.init_model

EstimatorEDncem.init_model(optimizer: str = 'adam', learning_rate: float = 0.0001, latent_dim: int = 10, dropout_rate: float = 0.1, l2_coef: float = 0.0, l1_coef: float = 0.0, enc_intermediate_dim: int = 128, enc_depth: int = 2, dec_intermediate_dim: int = 128, dec_depth: int = 2, cond_depth: int = 1, cond_dim: int = 8, cond_dropout_rate: float = 0.1, cond_activation: str = 'relu', cond_l2_reg: float = 0.0, cond_use_bias: bool = False, n_eval_nodes_per_graph: int = 32, use_domain: bool = False, scale_node_size: bool = True, beta: float = 0.01, max_beta: float = 1.0, pre_warm_up: int = 0, output_layer: str = 'gaussian', **kwargs)

Initialize a ModelEDncem object.

Parameters

• optimizer (str) – Optimizer.

• learning_rate (float) – Learning rate.

• latent_dim (int) – Latent dimension.

• dropout_rate (float) – Dropout.

• l2_coef (float) – l2 regularization coefficient.

• l1_coef (float) – l1 regularization coefficient.

• enc_intermediate_dim (int) – Encoder intermediate dimension.

• enc_depth (int) – Encoder depth.

• dec_intermediate_dim (int) – Decoder intermediate dimension.

• dec_depth (int) – Decoder depth.

• cond_depth (int) – Graph conditional depth.

• cond_dim (int) – Graph conditional dimension.

• cond_dropout_rate (float) – Graph conditional dropout rate.

• cond_activation (str) – Graph conditional activation.

• cond_l2_reg (float) – Graph conditional l2 regularization coefficient.

• cond_use_bias (bool) – Graph conditional use bias.

• n_eval_nodes_per_graph (int) – Number of nodes per graph.

• use_domain (bool) – Whether to use domain information.

• scale_node_size (bool) – Whether to scale output layer by node sizes.

• beta (float) – Beta used in BetaScheduler.

• max_beta (float) – Maximal beta used in BetaScheduler.

• pre_warm_up (int) – Number of epochs in pre warm up.

• output_layer (str) – Output layer.

• kwargs – Arbitrary keyword arguments.