ncem.estimators.Estimator

class ncem.estimators.Estimator[source]

Estimator class for models.

Contains all necessary methods for data loading, model initialization, training, evaluation and prediction.

Attributes

`img_keys_all`	Return all image keys.
`nodes_idx_all`	Return all node indices.
`nodes_idx_eval`	alias of `Dict`[`str`, `list`]
`nodes_idx_test`	alias of `Dict`[`str`, `list`]
`nodes_idx_train`	alias of `Dict`[`str`, `list`]
`patient_ids_bytarget`	Return patient identifiers by target.
`patient_ids_unique`	Return unique patient identifiers.
`img_to_patient_dict`
`complete_img_keys`
`a`
`h_0`
`h_1`
`size_factors`
`graph_covar`
`node_covar`
`domains`
`covar_selection`
`node_types`
`node_type_names`
`graph_covar_names`
`node_feature_names`
`n_features_type`
`n_features_standard`
`n_features_0`
`n_features_1`
`n_graph_covariates`
`n_node_covariates`
`n_domains`
`max_nodes`
`n_eval_nodes_per_graph`
`vi_model`
`log_transform`
`model_type`
`adj_type`
`cond_type`
`cond_depth`
`output_layer`
`steps_per_epoch`
`validation_steps`

Methods

`evaluate_any`(img_keys, node_idx[, batch_size])	Evaluate model on any given data set.
`evaluate_per_node_type`([batch_size])	Evaluate model for each node type seperately.
`get_data`(data_origin, data_path, radius[, ...])	Get data used in estimator classes.
`init_model`(**kwargs)	Initialize and compiles the model.
`predict`([batch_size])	Return observed labels and full predictions (including scale model) grouped exactly as in nodes_idx_test.
`pretrain_decoder`([decoder_epochs, patience, ...])	Pre-train decoder model.
`split_data_given`(img_keys_test, ...)	Split data by given partition.
`split_data_node`(test_split, validation_split)	Split nodes randomly into partitions.
`split_data_target_cell`(target_cell, ...[, seed])	Split nodes randomly into partitions.
`train`([epochs, epochs_warmup, ...])	Train model.
`train_aggressive`([aggressive_enc_patience, ...])	Train model aggressive.
`train_normal`([epochs, patience, ...])	Train model normal.