Loss Functions for Top-k Error: Analysis and Insights
This work addresses the need for better performance measures in computer vision tasks with large, ambiguous class sets, offering incremental improvements in top-k error optimization.
The paper tackles the problem of evaluating and improving top-k error in multiclass classification, especially for datasets with many ambiguous classes, by introducing novel top-k loss functions and efficient optimization schemes. The results show that the softmax loss is competitive across all k, while the new top-k losses provide further improvements for specific k values and train faster.
In order to push the performance on realistic computer vision tasks, the number of classes in modern benchmark datasets has significantly increased in recent years. This increase in the number of classes comes along with increased ambiguity between the class labels, raising the question if top-1 error is the right performance measure. In this paper, we provide an extensive comparison and evaluation of established multiclass methods comparing their top-k performance both from a practical as well as from a theoretical perspective. Moreover, we introduce novel top-k loss functions as modifications of the softmax and the multiclass SVM losses and provide efficient optimization schemes for them. In the experiments, we compare on various datasets all of the proposed and established methods for top-k error optimization. An interesting insight of this paper is that the softmax loss yields competitive top-k performance for all k simultaneously. For a specific top-k error, our new top-k losses lead typically to further improvements while being faster to train than the softmax.