Enhancing Personalized Ranking With Differentiable Group AUC Optimization
This work addresses the problem of optimizing AUC directly in binary classifiers for personalized ranking, particularly in recommendation systems, though it is incremental as it builds on existing pairwise loss methods.
The paper tackles the gap between training classifiers with cross-entropy and evaluating them with AUC by proposing PDAOM loss, a personalized differentiable AUC optimization method that improves offline AUC/GAUC metrics and reduces computational complexity, and achieves a 1.40% increase in clicks and 0.65% increase in orders in an online recommendation system.
AUC is a common metric for evaluating the performance of a classifier. However, most classifiers are trained with cross entropy, and it does not optimize the AUC metric directly, which leaves a gap between the training and evaluation stage. In this paper, we propose the PDAOM loss, a Personalized and Differentiable AUC Optimization method with Maximum violation, which can be directly applied when training a binary classifier and optimized with gradient-based methods. Specifically, we construct the pairwise exponential loss with difficult pair of positive and negative samples within sub-batches grouped by user ID, aiming to guide the classifier to pay attention to the relation between hard-distinguished pairs of opposite samples from the perspective of independent users. Compared to the origin form of pairwise exponential loss, the proposed PDAOM loss not only improves the AUC and GAUC metrics in the offline evaluation, but also reduces the computation complexity of the training objective. Furthermore, online evaluation of the PDAOM loss on the 'Guess What You Like' feed recommendation application in Meituan manifests 1.40% increase in click count and 0.65% increase in order count compared to the baseline model, which is a significant improvement in this well-developed online life service recommendation system.