tensorflow/adanet

Implements adaptive ensemble learning by iteratively growing neural network subnetworks while freezing previous iterations, optimizing an objective with theoretical learning guarantees based on the AdaNet algorithm. Integrates with TensorFlow's Estimator API and supports multi-task learning (regression, classification), distributed training across CPU/GPU/TPU, and extensible subnetwork definitions via `tf.layers`. Also provides `AutoEnsembleEstimator` for learning optimal ensembles from user-defined heterogeneous models.

3,457 stars and 291 monthly downloads. No commits in the last 6 months. Available on PyPI.