The PyTorch implementation of variants of loss for text classification & text matching.

基于PyTorch实现的文本分类与文本匹配损失函数变种

• Distilling the Knowledge in a Neural Network (NIPS 2014 DeepLearning Workshop) [paper] - Soft Target & Soft Softmax Loss
• FaceNet: A Unified Embedding for Face Recognition and Clustering (CVPR 2015) [paper] - Triplet Loss
• Applying Deep Learning to Answer Selection: A Study and An Open Task (ASRU 2015) [paper]
• Holistically-Nested Edge Detection (ICCV 2015) [paper] - Weigted Softmax Loss
• V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation (3DV 2016) [paper] - Dice Loss
• UnitBox: An Advanced Object Detection Network (ACM Multimedia 2016) [paper] - IoU Loss
• Rethinking the Inception Architecture for Computer Vision (CVPR 2016) [paper] - Label Smoothing
• A Discriminative Feature Learning Approach for Deep Face Recognition (ECCV 2016) [paper] - Center Loss
• Large-Margin Softmax Loss for Convolutional Neural Networks (ICML 2016) [paper] - L-softmax Loss
• SphereFace: Deep Hypersphere Embedding for Face Recognition (CVPR 2017) [paper] - A-softmax Loss
• Focal Loss for Dense Object Detection (ICCV 2017) [paper] - Focal Loss
• The Lovász-Softmax Loss: A Tractable Surrogate for The Optimization of The Intersection-over-Union Measure in Neural Networks (CVPR 2018) [paper] [code] - Lovasz Softmax Loss
• Island Loss for Learning Discriminative Features in Facial Expression Recognition (FG 2018) [paper] - Island Loss
• Feature Incay for Representation Regularization (ICLR 2018 Workshop) [paper]
• Additive Margin Softmax for Face Verification (ICLR 2018 Workshop) [paper] [code] - AM-softmax Loss
• AnatomyNet: Deep Learning for Fast and Fully Automated Whole-volume Segmentation of Head and Neck Anatomy (Medical Physics 2018) [paper] - Exponential Logarithmic Loss (Focal Loss + Dice Loss)
• ArcFace: Additive Angular Margin Loss for Deep Face Recognition (CVPR 2019) [paper] - AA-softmax Loss
• Mixtape: Breaking the Softmax Bottleneck Efficiently (NeurIPS 2019) [paper] - Mixtape
• When Does Label Smoothing Help? (NeuIPS 2019) [paper]
• Complement Objective Training (ICLR 2019) [paper] - COT
• Dice Loss for Data-imbalanced NLP Tasks (CoRR 2019) [paper] - Dice Loss
• Tversky Loss Function for Image Segmentation Using 3D Fully Convolutional Deep Networks (MLMI@MICCAI 2017) [paper] - Tversky Loss

• label smoothing 如何让正确类与错误类在 logit 维度拉远的？
• 如何理解soft target这一做法？ - 知乎
• 有哪些「魔改」loss函数，曾经拯救了你的深度学习模型？ - 知乎
• 图像分割中的loss--处理数据极度不均衡的状况
• 【损失函数合集】超详细的语义分割中的Loss大盘点 - 言有三
• 【技术综述】一文道尽softmax loss及其变种 - 言有三
• Softmax理解之从最优化的角度看待Softmax损失函数 - 王峰
• Softmax理解之二分类与多分类 - 王峰
• Softmax理解之Smooth程度控制 - 王峰
• Softmax理解之margin - 王峰
• Softmax理解之margin的自动化设置 - 王峰
• Softmax理解之被忽略的Focal Loss变种 - 王峰
• 文本情感分类（四）：更好的损失函数 - 科学空间
• 从loss的硬截断、软化到focal loss - 科学空间
• Keras中自定义复杂的loss函数 - 科学空间
• 基于GRU和am-softmax的句子相似度模型 - 科学空间