Vectorized Huber Loss and Gradient
Batch MSE and MAE with Gradients
Best Constant Prediction Under Regression Losses
Vectorized Quantile Pinball Loss
Numerically Stable Log-Cosh Loss and Gradient
Margin-Based Classification Surrogate Losses
Multiclass SVM Hinge Loss and Subgradient
Cost-Sensitive Binary Cross-Entropy from Logits
Weighted Multi-Label BCE From Logits
Label-Smoothed Cross-Entropy with Ignore Index
MSE-on-Sigmoid vs BCE Logit Gradients
Focal Loss From Logits With Gradients
Metric Learning Contrastive Losses
Contrastive Loss for Siamese Pairs
Cosine Embedding Loss With Safe Norms
NT-Xent Contrastive Loss and Gradient
Triplet Loss with Semi-Hard Negative Mining
Probabilistic Likelihood Losses
KL Divergence Loss with Zero-Probability Targets
Poisson NLL Loss From Log-Rates
Heteroscedastic Gaussian NLL with Clamped Log-Variance
Tweedie Deviance Loss With Log-Link Gradient
Specialized Composite Losses
Pairwise Margin Ranking Loss
Weighted Loss Wrapper with Reduction Modes
Knowledge Distillation Loss and Gradient
Vectorized RankNet Loss and Gradient
Soft Dice and BCE Hybrid Segmentation Loss
Uncertainty-Weighted Multi-Task Loss
Guarded BCE Loss with Clipping and NaN Detection
DPO Loss and Gradient in NumPy
Asymmetric Pseudo-Huber Boosting Objective
ListNet Top-One Loss with Padding Masks