Skip to content

python-package.md

Latest commit

 

History

History
81 lines (59 loc) · 3.49 KB

python-package.md

File metadata and controls

81 lines (59 loc) · 3.49 KB
Raw
description Using the MED3pa library directly, outside the application.

🐍 The MED3pa package

Everything the application computes is computed by the MED3pa Python package. The interface exists to make the method reachable without writing code; the package is there when you need to go further: batching many experiments, embedding the analysis in a pipeline, or extending the method itself.

The med3pa subpackage: IPC, APC and MPC models and the tree representation behind the profiles

Subpackages

Subpackage Responsibility
datasets Stores and manages the dataset
models Handles ML model operations, including the base model wrapper
med3pa Evaluates the model's performance and extracts disadvantaged profiles

Installation

pip install MED3pa

{% hint style="warning" %} That installs the current stable release. The application pins a pre-release, MED3pa==1.1.0a3, which pip will not resolve to unless you ask for it by version or pass --pre. To match what the app runs:

pip install MED3pa==1.1.0a3

{% endhint %}

The package declares Requires-Python >=3.9; the application bundles 3.12. Pinning the surrounding dependencies matters as much as the version itself, so see Quick start for the combination that resolves.

A simple example

from MED3pa.datasets import DatasetsManager
from MED3pa.med3pa import Med3paExperiment
from MED3pa.models import BaseModelManager
from MED3pa.visualization.mdr_visualization import visualize_mdr
from MED3pa.visualization.profiles_visualization import visualize_tree

# Initialize the DatasetsManager
datasets = DatasetsManager()
datasets.set_from_data(dataset_type="testing",
                       observations=x_evaluation.to_numpy(),
                       true_labels=y_evaluation,
                       column_labels=x_evaluation.columns)

# Initialize the BaseModelManager
base_model_manager = BaseModelManager(model=clf)

# Execute the MED3pa experiment
results = Med3paExperiment.run(
    datasets_manager=datasets,
    base_model_manager=base_model_manager,
    **med3pa_params
)

# Save the results to a specified directory
results.save(file_path='results/oym')

# Visualize results
visualize_mdr(result=results, filename='results/oym/mdr')
visualize_tree(result=results, filename='results/oym/profiles')

The med3pa_params dictionary holds the same settings the Configuration page fills in: the IPC and APC hyperparameters, the MPC strategy and the samples-ratio sweep.

{% hint style="info" %} Passing a built-in confidence metric by name is known to raise a TypeError in the library. Pass the metric callable instead if you hit it. The application always passes callables, which is why the same configuration works there. {% endhint %}

Going further

  • Package documentation: tutorials for the datasets, models and med3pa subpackages.
  • Example: runnable notebooks, including a full one-year-mortality study.
  • study_3pa: the complete code behind the results reported in the JAMIA article.

Please feel free to contact us if you need any further assistance 😇.