Operations

When creating an agreement to allow a counterparty to make use of your asset in model training, use add_agreement() with the appropriate tripleblind.Operation value for the operation parameter as specified in the table below. For example, to create an agreement for Blind Join, use tripleblind.Operation.BLIND_JOIN.

When creating an agreement to allow a counterparty to infer against your trained model, use tripleblind.Operation.EXECUTE for the operation parameter.

Operation Agreement How to use
Blind Join BLIND_JOIN

⚠️Permissive agreements (without usage restrictions) are not recommended for this operation.

Use the blind_join() method to join your dataset with one or more dataset assets.
Neural Networks BLIND_LEARNING Use create_job() to train a model, with BLIND_LEARNING for the operation parameter.
Neural Networks: Split Datasets VERTICAL_BLIND_LEARNING Use create_job() to train a model, with VERTICAL_BLIND_LEARNING for the operation parameter. ℹ️Vertically-partitioned records must be in corresponding order across all data sources.
Neural Networks: Vertical Datasets PSI_VERTICAL_BLIND_LEARNING Use create_job() to train a model, with PSI_VERTICAL_BLIND_LEARNING for the operation parameter.
Blind Match PRIVATE_SET_INTERSECTION Use the intersect() method to carry out a private set intersection between your dataset and one or more additional dataset assets.
Blind Query BLIND_QUERY

⚠️Permissive agreements (without usage restrictions) are not recommended for this operation.

Use create_job() to query a dataset asset, with BLIND_QUERY for the operation parameter.
Blind Report When using add_agreement() to allow a counterparty to run your Blind Report, use EXECUTE for the operation parameter. Use create_job() to run the report in a process, with the positioned ReportAsset‘s UUID for the operation parameter.
Blind Sample BLIND_SAMPLE Use the get_sample() method to generate a realistic synthetic sample for a dataset asset.
Blind Stats STATS Use the get_statistics() method to query descriptive statistics against one or more dataset assets.
Blind String Search REGEX_COUNT Use create_job() to obtain counts of text search matches, with REGEX_COUNT for the operation parameter.
Decision Tree: Vertical Datasets PSI_VERTICAL_DECISION_TREE_TRAIN Use create_job() to train a model, with PSI_VERTICAL_DECISION_TREE_TRAIN for the operation parameter.
K-Means Clustering: Vertical Datasets PSI_VERTICAL_KMEANS_TRAIN Use create_job() to train a clustering model, with PSI_VERTICAL_KMEANS_TRAIN for the operation parameter.
NLP - BERT BERT_SEQ_CLF_TRAIN Use create_job() to train your model, with BERT_SEQ_CLF_TRAIN for the operation parameter.
Natural Language Processing NLP_TRAIN Use create_job() to train your model, with NLP_TRAIN for the operation parameter.
Outlier Detection OUTLIER_DETECTION Use the detect_outlier() method to identify the indices of outlier values in your dataset asset.
Random Forest RANDOM_FOREST_TRAIN Use create_job() to train a model, with RANDOM_FOREST_TRAIN for the operation parameter.
Recommender Model RECOMMENDER_TRAIN Use create_job() to train a Recommender Model, with RECOMMENDER_TRAIN for the operation parameter.
Regression REGRESSION For training, use the RegressionModel.train() method to train a horizontally-partitioned Linear, Logistic, Lasso, or Ridge regression model. For inference, use the ModelAsset.infer() method to infer on a trained model.
Regression: Vertical Datasets PSI_VERTICAL_REGRESSION_TRAIN For training, use the PSIVerticalRegressionModel.train() method to identify an overlap of matching records across datasets, and train a linear or logistic regression model on the vertically-partitioned intersection. For inference, use the ModelAsset.psi_infer() method to infer on a trained model.
XGBoost XGBOOST_TRAIN Use the XGBoost train(), predict(), and predict_proba() methods.