---
# name: "kg-phenio"
# description: "kg-phenio"
Target:
  target_path: graph_ml

Upload:
  s3_bucket: kg-hub-public-data
  s3_bucket_dir: kg-phenio/20240111/graph_ml/
  extra_args:
    "ACL": "public-read"

GraphDataConfiguration:
  source_data:
    files:
      - path: https://kg-hub.berkeleybop.io/kg-phenio/20240111/kg-phenio.tar.gz
        desc: "Location of KG-Phenio nodefile, edgefile, prefixcats nodes, and validation subgraphs."
  graph:
    directed: False
    node_path: merged-kg_nodes-prefixcats.tsv
    edge_path: merged-kg_edges.tsv
    verbose: True
    nodes_column: 'id'
    node_list_node_types_column: 'category'
    default_node_type: 'biolink:NamedThing'
    sources_column: 'subject'
    destinations_column: 'object'
    default_edge_type: 'biolink:related_to'
  evaluation_data:
    valid_data:
      pos_edge_filepath: pos_valid_edges.tsv
      neg_edge_filepath: neg_valid_edges.tsv
    train_data:
      neg_edge_filepath: neg_train_edges.tsv

# Generate new embeddings
# These stats are reduced from usual while I sort out some bugs
EmbeddingsConfig:
  filename: kg-phenio-SkipGram-minimal.tsv
  history_filename: embedding_history.json
  node_embedding_params:
    node_embedding_method_name: SkipGram
    use_mirrored_strategy: False
    walk_length: 20 #100
    batch_size: 64 #128
    window_size: 4
    return_weight: 1.0
    explore_weight: 1.0
    iterations: 5 #20
  tsne_file_name: tsne.png

ClassifierContainer:
  classifiers:
  - classifier_id: mlp_0
    classifier_name: neural network
    classifier_type: tensorflow.keras.models.Sequential
    edge_method: Average
    outfile: "model_mlp_kg-phenio.model"
    history_filename: "model_mlp_kg-phenio_history.json"
    parameters:
      tf_keras_params:
        layers_config:
          layers:
            - type: tensorflow.keras.layers.Input
              parameters:
                shape: 100   # must match embedding_size up above
            - type: tensorflow.keras.layers.Dense
              parameters:
                units: 128
                activation: relu
            - type: tensorflow.keras.layers.Dense
              parameters:
                units: 32
                activation: relu
            - type: tensorflow.keras.layers.Dropout
              parameters:
                rate: 0.5
            - type: tensorflow.keras.layers.Dense
              parameters:
                units: 16
                activation: relu
            - type: tensorflow.keras.layers.Dense
              parameters:
                units: 1
                activation: sigmoid
        loss: binary_crossentropy
        optimizer: nadam
        metrics_config:
          metrics:
            - name: auprc
              type: tensorflow.keras.metrics.AUC
              curve: PR
            - name: auroc
              type: tensorflow.keras.metrics.AUC
              curve: ROC
            - name: Recall
              type: tensorflow.keras.metrics.Recall
            - name: Precision
              type: tensorflow.keras.metrics.Precision
            - type: accuracy
        fit_config:
          batch_size: 4096
          epochs: 10
          callbacks_list:
            callbacks:
              - type: tensorflow.keras.callbacks.EarlyStopping
                monitor: val_loss
                patience: 5
                min_delta: 0.001  
              - type: tensorflow.keras.callbacks.ReduceLROnPlateau
  # - classifier_id: dt_0
  #   type: Decision Tree
  #   edge_method: Average
  #   model:
  #     outfile: "model_decisiontree_kg-ontoml.model"
  #     type: sklearn.tree.DecisionTreeClassifier
  #     parameters:
  #       max_depth: 30
  #       random_state: 42
  - classifier_id: rf_0
    classifier_name: Random Forest
    classifier_type: sklearn.ensemble.RandomForestClassifier
    edge_method: Average
    outfile: "model_randomforest_kg-phenio.model"
    parameters:
      sklearn_params:
        n_estimators: 500
        max_depth: 30
        n_jobs: 8
        random_state: 42
  - classifier_id: lr_0
    classifier_name: Logistic Regression
    classifier_type: sklearn.linear_model.LogisticRegression
    edge_method: Average
    outfile: "model_lr_kg-phenio.model"
    parameters:
      sklearn_params:
        random_state: 42
        max_iter: 2000

ApplyTrainedModelsContainer:
  models:
    - model_id: mlp_0
      cutoff: 0.9
      outfile: mlp_classifier_predictions_kgx.tsv
      node_types:
        source:
          - "HP"
        destination:
          - "MP"
  # -
  #   classifier_model_id: dt_0
  #   cutoff: 0.9
  #   outfile: dt_classifier_predictions_kgx.tsv
  #   link_node_types:
  #     source:
  #       - 'HP'
  #     destination:
  #       - 'MP'
    - model_id: rf_0
      cutoff: 0.9
      outfile: rf_classifier_predictions_kgx.tsv
      node_types:
        source:
          - "HP"
        destination:
          - "MP"
    - model_id: lr_0
      cutoff: 0.9
      outfile: lr_classifier_predictions_kgx.tsv
      node_types:
        source:
          - "HP"
        destination:
          - "MP"