This python modules contains a collection of custom Tensorflow Keras objects for:
- hierarchical multilabel learning (TMLoss, MCLoss, hierarchical regularization) and inference (hierarchical roll-up and roll-down)
- cost sensitive label imbalance correction
- recommender system performance evaluation (NDCG, RE@R)
- handling of evaluation of more than one dev/test set during training
- manipulation of dense and sparse tensors