bernn 0.4.4

Batch Effect Removal Neural Networks for Tandem Mass Spectrometry

BERNN-MSMS

Minimal README for quick usage.

Longer historical content is kept in LEGACY_README.md.

Install

pip install bernn

Basic usage

from bernn import TrainAEClassifierHoldout

trainer_cls = TrainAEClassifierHoldout
trainer = trainer_cls(config=bernn_config, log_metrics=True, keep_models=False)

# Train and predict in one call
preds_encoded = trainer.fit_predict(
    X_train,
    y_train,
    X_test=X_test,
    y_test=y_test,
    groups_train=batches_train,
    groups_test=batches_test,
    cross_validation=False,
    cross_test=False,
)

# Decode predictions back to original labels
preds = trainer.predict(X_test)

Important runtime contract:

• groups_train is mandatory.
• If X_test is provided, groups_test is mandatory.

Important parameters

Focus on these first:

• optimize_hyperparams: enable/disable Ax optimization.
• n_trials: number of optimization trials.
• fixed_hyperparams: force values and remove them from search.
• n_repeats: number of holdout repeats.
• n_layers, layer1: classifier depth and width seed.
• dloss: domain loss mode.
• warmup, n_epochs: core training schedule.
• device: cpu/cuda target.
• scaler, bs: preprocessing and batch size.

Official documentation

• Full reference: OFFICIAL_DOCUMENTATION.md
• Full parameter catalog: TRAINING_PARAMETERS.md
• Minimal runnable examples notebook (4 variants): tutorials/minimal_examples.ipynb
• Optimized all-config notebook (TrainAEClassifierHoldout): tutorials/optimized_classifier_holdout_all_configs.ipynb
• Optimized all-config notebook (TrainAEThenClassifierHoldout): tutorials/optimized_ae_then_classifier_holdout_all_configs.ipynb
• Historical CLI-heavy guide: LEGACY_README.md