Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Distributional Loss for Robust Classification

About

This paper proposes a novel loss concept for supervised classification tasks. Rather than enforcing a direct mapping from each input sample to a single assigned label, we define an optimization objective over all classifier outputs as a bimodal Gaussian distribution. This softer target formulation implicitly captures class ambiguity, mitigates overfitting, and encourages the learning of more robust decision boundaries, all without requiring additional label information. Experimental results demonstrate consistent improvements in robustness, with particularly pronounced gains in low-data regimes, while requiring only minimal modifications to standard training pipelines.

Kathleen Anderson, Thomas Martinetz• 2026

Related benchmarks

TaskDatasetResultRank
Image ClassificationCaltech-256 (test)
Top-1 Acc80.91
86
Image ClassificationBloodMNIST (test)
Accuracy96.71
34
Image ClassificationPathMNIST (test)
Accuracy@1 (%)99.61
20
Showing 3 of 3 rows

Other info

Follow for update