pyltr

pyltr is a Python learning-to-rank toolkit with ranking models, evaluation metrics, data wrangling helpers, and more.

This software is licensed under the BSD 3-clause license (see LICENSE.txt).

Example

Import pyltr:

import pyltr

Import a LETOR dataset (e.g. MQ2007 ):

with open('train.txt') as trainfile, \
        open('vali.txt') as valifile, \
        open('test.txt') as evalfile:
    TX, Ty, Tqids, _ = pyltr.data.letor.read_dataset(trainfile)
    VX, Vy, Vqids, _ = pyltr.data.letor.read_dataset(valifile)
    EX, Ey, Eqids, _ = pyltr.data.letor.read_dataset(evalfile)

Train a LambdaMART model, using validation set for early stopping and trimming:

metric = pyltr.metrics.NDCG(k=10)

# Only needed if you want to perform validation (early stopping & trimming)
monitor = pyltr.models.monitors.ValidationMonitor(
    VX, Vy, Vqids, metric=metric, stop_after=250)

model = pyltr.models.LambdaMART(
    metric=metric,
    n_estimators=1000,
    learning_rate=0.02,
    max_features=0.5,
    query_subsample=0.5,
    max_leaf_nodes=10,
    min_samples_leaf=64,
    verbose=1,
)

model.fit(TX, Ty, Tqids, monitor=monitor)

Evaluate model on test data:

Epred = model.predict(EX)
print 'Random ranking:', metric.calc_mean_random(Eqids, Ey)
print 'Our model:', metric.calc_mean(Eqids, Ey, Epred)

Features

Below are some of the features currently implemented in pyltr.

Models

• LambdaMART (pyltr.models.LambdaMART)
  • Validation & early stopping
  • Query subsampling

Metrics

• (N)DCG (pyltr.metrics.DCG, pyltr.metrics.NDCG)
  • pow2 and identity gain functions
• ERR (pyltr.metrics.ERR)
  • pow2 and identity gain functions
• (M)AP (pyltr.metrics.AP)
• Kendall's Tau (pyltr.metrics.KendallTau)
• AUC-ROC -- Area under the ROC curve (pyltr.metrics.AUCROC)

Data Wrangling

• Data loaders (e.g. pyltr.data.letor.read)
• Query groupers and validators (pyltr.util.group.check_qids, pyltr.util.group.get_groups)

Running Tests

Use the run_tests.sh script to run all unit tests.

Building Docs

cd into the docs/ directory and run make html. Docs are generated in the docs/_build directory.