A no-regret generalization of hierarchical softmax to extreme multi-label classification
This addresses the problem of efficiently tagging instances with relevant labels from extremely large pools for applications like online systems, though it is incremental as it builds on existing probabilistic label trees.
The paper tackles extreme multi-label classification by showing that probabilistic label trees (PLTs) are a no-regret generalization of hierarchical softmax, proving the inconsistency of the pick-one-label heuristic, and demonstrating that their implementation (extremeText) achieves significantly better results than hierarchical softmax with pick-one-label and XML-CNN, while being competitive with state-of-the-art approaches in statistical performance, model size, and prediction time.
Extreme multi-label classification (XMLC) is a problem of tagging an instance with a small subset of relevant labels chosen from an extremely large pool of possible labels. Large label spaces can be efficiently handled by organizing labels as a tree, like in the hierarchical softmax (HSM) approach commonly used for multi-class problems. In this paper, we investigate probabilistic label trees (PLTs) that have been recently devised for tackling XMLC problems. We show that PLTs are a no-regret multi-label generalization of HSM when precision@k is used as a model evaluation metric. Critically, we prove that pick-one-label heuristic - a reduction technique from multi-label to multi-class that is routinely used along with HSM - is not consistent in general. We also show that our implementation of PLTs, referred to as extremeText (XT), obtains significantly better results than HSM with the pick-one-label heuristic and XML-CNN, a deep network specifically designed for XMLC problems. Moreover, XT is competitive to many state-of-the-art approaches in terms of statistical performance, model size and prediction time which makes it amenable to deploy in an online system.