深度学习论文
蜻蜓点论文， paperskim, deep learning papers. the table of content of all my videos
- 方法类的论文， 不保证方法理解的正确和准确
- 架构和方法类的论文， 不深挖背后的数学**
- 实验部分 或 实验类的论文， 基本上简单翻译
如有问题， 欢迎提问；如有质疑， 欢迎右上（手机好像是左下和左上？）
B站Think不Clear Youtube(PaperThinkNotClear)
西瓜视频 因标题字数限制，我基本都改成中文名了， 所以， 我自己都没法搜索到目标视频
幻灯片 Slides on OneDrive 没有同步， 因为需要同时存 百度和onedrive， 怎么存呢？ 只能是懒得管了
百度网盘 baidu pan 我搞错了
链接 https://pan.baidu.com/s/1e3lh08SE6mKg3E7loktQUA 提取码：e0gn
同样可用 链接：https://pan.baidu.com/s/1fTQnIGhQ3hcvjlDrM4NNFA 提取码：ks3c
| 论文名 | Bilibili | Youtube | Arxiv | 博客 | 序号 |
|---|---|---|---|---|---|
| 239 | |||||
| 238 | |||||
| 237 | |||||
| StyleGAN2: Analyzing and Improving the Image Quality of StyleGAN | 236 | ||||
| DatasetGAN: Efficient Labeled Data Factory with Minimal Human Effort | 235 | ||||
| StyleGAN: A Style-Based Generator Architecture for GAN | 234 | ||||
| AdaIN:Arbitrary Style Transfer in Real-time with Adaptive Instance Norm | 233 | ||||
| A learned representation for artistic style | 232 | ||||
| Cold Diffusion Inverting Arbitrary Image Transforms Without Noise | 231 | ||||
| EnergyMatch: Energy-based Pseudo-Labeling for Semi-Supervised | 230 | ||||
| Knowledge distillation A good teacher is patient and consistent | 229 | ||||
| Understanding Attention for Vision-and-Language Tasks | 228 | ||||
| Efficient Training of Visual Transformers with Small Datasets | 227 | ||||
| Convolutional Knowledge Tracing Individualization in Learning Process | 226 | ||||
| Educational Question Mining At Scale Prediction, Analysis and Personal | 225 | ||||
| EDDI Dynamic Discovery of High-Value Information with Partial VAE | 224 | ||||
| Diffusion Models Beat GANs on Image Synthesis | 223 | ||||
| Accelerating Diffusion Models via Early Stop of the Diffusion Process | 222 | ||||
| Improved Denoising Diffusion Probabilistic Models | 221 | ||||
| UserBERT contrastive and self-supervised | 220 | ||||
| An Empirical Study of Train End-to-End Vision-and-Language Transformer | 219 | ||||
| 多模态预训练 VLMo 和 VL-BEIT | 218 | ||||
| Unsupervised Vision-and-Language Pre-train wo Parallel Images captions | 217 | ||||
| ViLT Vision-and-Language Transformer Without Convolution or Region | 216 | ||||
| Parameter-Efficient Transfer Learning for NLP | 215 | ||||
| Sharpness-Aware Training for Free | 214 | ||||
| When Vision Transformers Outperform ResNets wo Pre-training Strong DA | 213 | ||||
| SLViT: Vision Transformer for Small-Size Datasets | 212 | ||||
| CCT: Escaping the Big Data Paradigm with Compact Transformers | 211 | ||||
| Self-supervised Graph Learning for Recommendation | 210 | ||||
| LightGCN Simplifying Graph Convolution Network for Recommendation | 209 | ||||
| MiniRocket: Very Fast Deterministic Time Series Classification | 208 | ||||
| Simplifying Graph Convolutional Networks | 207 | ||||
| Deep Generative prior for Image Restoration and Manipulation | 206 | ||||
| GCN semi-supervised classification with graph conv networks | youtube | 205 | |||
| Improved Trainable Calibration Method on Medical Imaging Classification | 204 | ||||
| Graph Attention Networks | 203 | ||||
| Anomaly Transformer: Time Series Anomaly Detection | 202 | ||||
| [RandAugument and Cutout](https://www.bilibili.com/video/BV1V5411d7D3"【凑个数】201 RandAugument and Cutout") | 201 | ||||
| Crafting Better Contrastive Views for Siamese Representation Learning | 200 | ||||
| Efficient Sharpness-aware Minimization for Improve Neural Networks | 198 | ||||
| Improved Contrastive Divergence Training of Energy-Based Model | 199 | ||||
| The Effects of Regularization and Data Augmentation are Class Dependent | 197 | ||||
| Tradeoffs in Data Augmentation An Empirical Study | 196 | ||||
| Visual Prompting: Modifying Pixel Space to Adapt Pre-trained Models | 195 | ||||
| SPICE Semantic Pseudo-Labeling for Image Clustering | 194 | ||||
| Sharpness-Aware Minimization for Efficiently Improving Generalization | 194 | ||||
| DeCLUTR Contrastive Learning for Unsupervised Textual Representation | 192 | ||||
| Revisiting the Transferability of Supervised Pretraining via an MLP | 191 | ||||
| CoMatch: Semi-supervised Learning with Contrastive Graph Regularization | 190 | ||||
| Event Extraction by Answering (Almost) Natural Questions | 189 | ||||
| You never cluster Alone | You Never Cluster Alone 1.1 | 188 | |||
| Complement Objective Training | 187 | ||||
| Well-classified Examples are Underestimated in Classification with DNN | 186 | ||||
| When Does Label Smoothing Help | 185 | ||||
| SEED: Self-supervised Distillation For Visual Representation | 184 | ||||
| MixText: Hidden Space MixUp for Semi-Supervised Text Classification | 183 | ||||
| Nearest Neighbor Matching for Deep Clustering | 182 | ||||
| Conditional Self-Supervised Learning for Few-Shot Classification | 181 | ||||
| Active Learning at the ImageNet Scale | 180 | ||||
| Towards Understand Generative Capability Adversarial Robust Classifier | 179 | ||||
| FlexMatch Semi-Supervised Learning with Curriculum Pseudo Labeling | 178 | ||||
| Learning Energy-Based Models by Diffusion Recovery Likelihood | 177 | ||||
| Self-Knowledge Distillation with Progressive Refinement of Targets | 176 | ||||
| AEDA: An Easier Data Augmentation Technique for Text Classification | 175 | ||||
| VAEBM: Variational Autoencoders and Energy-based Models | 174 | ||||
| On Separability of Self-Supervised Representations | 173 | ||||
| Revisiting Knowledge Distillation via Label Smoothing Regularization | 172 | ||||
| Regularizing Class-wise Predictions via Self-knowledge Distillation | 171 | ||||
| Be Your Own Teacher: Improve CNN via Self Distillation | 170 | ||||
| 以下删除过 | |||||
| Bayesian Deep Learning and a Probabilistic Perspective of Generalization | 169 | ||||
| Rethink Image Mixture for Unsupervised Visual Representation Learning | 168 | ||||
| FixMatch: Semi-Supervised Learning with Consistency and Confidence | 167 | ||||
| ReMixMatch: Semi-Supervised Learning with Distribution Alignment and Augmentation Anchoring | 166 | ||||
| MiCE: Mixture of Contrastive Experts for Unsupervised Image Clustering | 165 | ||||
| Learn Representation via Information Maximizing Self-Augmented Training | 164 | ||||
| Supporting Clustering with Contrastive Learning | 163 | ||||
| Unsupervised Multi-hop Question Answering by Question Generation | 162 | ||||
| Perceiver: General Perception with Iterative Attention | 161 | ||||
| Joint EBM Training for Better Calibrated NLU Models | 160 | ||||
| A Unified Energy-Based Framework for Unsupervised Learning | 159 | ||||
| Energy-Based Models for Deep Probabilistic Regression | 158 | ||||
| Contrastive Learning Inverts the Data Generating Process | 157 | ||||
| Asymmetric Loss For Multi-Label Classification | 156 | ||||
| Computation-Efficient Knowledge Distillation by Uncertainty-Aware Mixup | 155 | ||||
| Knowledge Distillation Meets Self-Supervision | 154 | ||||
| Feature Projection for Improved Text Classification | 153 | ||||
| Improve Joint Train of Inference Net and Structure Predict EnergyNet | 152 | ||||
| BertGCN: Transductive Text Classification by Combining GCN and Bert | 151 | ||||
| The Authors Matter Understand Mitigate Implicit Bias in text classification | 150 | ||||
| Learning Approximate Inference Networks for Structured Prediction | 149 | ||||
| End-to-End Learning for Structured Prediction Energy Networks | 148 | ||||
| Revisiting Unsupervised Relation Extraction | 147 | ||||
| Sentence Meta-Embeddings for Unsupervised Semantic Textual Similarity | 146 | ||||
| X-Class: Text Classification with Extremely Weak Supervision | 145 | ||||
| Paint by Word | 144 | ||||
| Shape-Texture Debiased Neural Network Training | 143 | ||||
| Contrastive Learning through Alignment and Uniformity on the Hypersphere | 142 | ||||
| Deep INFOMAX representation mutual information estimation maximization | 141 | ||||
| SimCSE: Simple Contrastive Learning of Sentence Embeddings | 140 | ||||
| IMOJIE Iterative Memory-Based Joint Open Information Extraction | 139 | ||||
| Trash is Treasure Resisting Adversarial Examples by Adversarial Examples | 138 | ||||
| Enhancing Adversarial Defense by k-Winners-Take-All | 137 | ||||
| On Adaptive Attacks to Adversarial Example Defenses | 136 | ||||
| Knowledge distillation via softmax regression representation learning | 135 | ||||
| Revisiting Locally Supervised Learning Alternative to End-to-end Training | 134 | ||||
| Putting An End to End-to-End Gradient-Isolated Learning of Representations | 133 | ||||
| 防御defense变分自编码器 | 132 | ||||
| Triple Wins Accuracy Robustness Efficiency by Input-adaptive Inference | 131 | ||||
| Using latent space regression to analyze leverage compositionality in GANs | 130 | ||||
| Theoretically(没看) Principled Trade-off between Robustness and Accuracy | 129 | ||||
| Representation learning with contrastive predictive coding | 128 | ||||
| Learning Representations for Time Series Clustering | 127 | ||||
| Stochastic Security: Adversarial Defense Using Long-Run Dynamics of EBM | 126 | ||||
| Improving Adversarial Robustness via Channel-wise Activation Suppressing | 125 | ||||
| Likelihood Landscapes: A Unifying Principle Behind Adversarial Defenses | 124 | ||||
| Barlow Twins: Self-Supervised Learning via Redundancy Reduction | 123 | ||||
| Geometry-Aware Instance-Reweighted Adversarial Training | 122 | ||||
| A Closer Look at Accuracy vs Robustness | 121 | ||||
| Unsupervised Clustering of Seismic Signals 地震波 using autoencoders | 120 | ||||
| Towards the first adversarially robust neural network model on MNIST | 119 | ||||
| PGD对抗训练 Towards Deep Learning Models Resistant to Adversarial Attacks | 118 | ||||
| Denoising Diffusion Probabilistic Models | 117 | ||||
| Deep Unsupervised Learning using Nonequilibrium Thermodynamics | 116 | ||||
| Variational Inference with Normalizing Flows | 115 | ||||
| CutMix Regularization Strategy with Localizable Features | 114 | ||||
| Clustering-friendly Representation Learning Feature Decorrelate | 113 | ||||
| Energy-based Out-of-distribution Detection | 112 | ||||
| High-Performance Large-Scale Image Recognition Without Normalization | 111 | ||||
| Characterizing signal propagation in unnormalized ResNets | 110 | ||||
| Concept Learners for Few-Shot Learning | 109 | ||||
| Image Generation by Minimize Frechet Distance in Discriminator feature space | 108 | ||||
| Learning Non-Convergent Non-Persistent Short-Run MCMC to EBM | 107 | ||||
| Concept Whitening for Interpretable Image Recognition | 106 | ||||
| Loss Landscape Sightseeing with Multi-Point Optimization | 105 | ||||
| Loss Surfaces, Mode Connectivity, and Fast Ensembling of DNNs | 104 | ||||
| Essentially No Barriers in Neural Network Energy Landscape | 103 | ||||
| Visualizing the Loss Landscape of Neural Nets | 102 | ||||
| Self-training for Few-shot Transfer Across Extreme Task Differences | 101 | ||||
| Darts: Differentiable architecture search | 100 | ||||
| Architecture Search Space in Neural Architecture Search(NAS) | 99 | ||||
| Free Lunch for Few-shot Learning: Distribution Calibration | 98 | ||||
| Online Deep Clustering for Unsupervised Representation Learning | 97 | ||||
| Coarse-to-Fine Pre-training for Named Entity Recognition | 96 | ||||
| Unsupervised Domain Adaptation with Variational Information Bottleneck | 95 | ||||
| A Unified MRC Framework for Named Entity Recognition | 94 | ||||
| Rethinking Softmax Cross-Entropy Loss for Adversarial Robustness | 93 | ||||
| Super-Convergence: Very Fast Training of NN use large LR | 92 | ||||
| Contrastive Clustering | 91 | ||||
| Graph Contrastive Learning with Augmentations NIPS 2020 / Graph Contrastive Learning with Adaptive Augmentation WWW 2021 GCC: Graph Contrastive Coding for Graph Neural Network ... KDD 2021 | GNN and Contrastive Learning | 90 | |||
| Contrastive Representation Distillation | 89 | ||||
| Spectral Norm Regularization for Improving the Generalizability of NN | 88 | ||||
| UDA: Unsupervised Data Augmentation for Consistency Training | 87 | ||||
| Simplify the Usage of Lexicon in Chinese NER | 86 | ||||
| Chinese NER Using Lattice LSTM | 85 | ||||
| Uncertainty-aware Self-training for Few-shot Text Classification | 84 | ||||
| Concept Learning with Energy-Based Models | 83 | ||||
| Training data-efficient image transformers distillation through attention | 82 | ||||
| Adversarial Training Methods for Semi-Supervised Text Classification | 81 | ||||
| Delta-training Semi-Supervised Text Classification with word embedding | 80 | ||||
| On the Anatomy of MCMC-Based Maximum Likelihood Learning of EBMs | 79 | ||||
| Unsupervised Deep Embedding for Clustering Analysis | 78 | ||||
| Relation of Relation Learning Network for Sentence Semantic Matching | 77 | ||||
| Contextual Parameter Generation for Universal Neural Machine Translation | 76 | ||||
| Exploring Simple Siamese Representation Learning | 75 | ||||
| Contextual Parameter Generation for Knowledge Graph Link Prediction | 74 | ||||
| Robustness May Be at Odds with Accuracy | 73 | ||||
| Learning with Multiplicative Perturbations | 72 | ||||
| When Do Curricula Work？ | 71 | ||||
| Self-Supervised Contrastive Learning with Adversarial Examples | 70 | ||||
| Supervised Contrastive Learning | 69 | ||||
| A Note on the Inception Score and FID | 68 | ||||
| Hierarchical Semantic Aggregation for Contrastive Representation Learning | 67 | ||||
| Syntactic and Semantic-driven Learning for Open Information Extraction | 66 | ||||
| Text Classification with Negative Supervision | 65 | ||||
| CESI Canonicalizing Open Knowledge Bases by Embeddings and Side Information | 64 | ||||
| CaRe: Open Knowledge Graph Embedding | 63 | ||||
| No MCMC for me, Amortized sampling for fast and stable training of EBMs | 62 | ||||
| Knowledge Graph Embedding Based Question Answering | 61 | ||||
| VAT Virtual Adversarial Training for regularization semi-supervised learn | 60 | ||||
| CNN-Generated Images Are Surprisingly Easy to Spot.. For Now | 59 | ||||
| Graph Agreement Models for Semi-Supervised Learning | 58 | ||||
| Be More with Less: Hypergraph Attention Networks for Inductive 文本分类 | 57 | ||||
| Text Level Graph Neural Network for Text Classification | 56 | ||||
| Graph Convolutional Networks for Text Classification | 55 | ||||
| Learning sparse neural networks through L0 regularization | 54 | ||||
| BigGAN: Large Scale GAN Training for High Fidelity Natural Image Synthesis | 53 | ||||
| On the steerability of generative adversarial networks | 52 | ||||
| What Makes for Good Views for Contrastive Learning | 51 | ||||
| Viewmaker Networks Learning Views for Unsupervised Representation Learning | 50 | ||||
| Auto-Encoding Variational Bayes | 49 | ||||
| Adversarial Examples Improve Image Recognition | 48 | ||||
| Stochastic Weight Averaging for Generalization | 47 | ||||
| There Are Many Consistent Explanations Of Unlabeled Data | 46 | ||||
| Interpretable Convolutional Neural Networks | 45 | ||||
| Understanding Black-box Predictions via Influence Functions | 44 | ||||
| Adversarial Examples Are Not Bugs, They Are Features | 43 | ||||
| You Only Propagate Once Accelerating AT via Maximal Principle | 42 | ||||
| Text Classification Using Label Names Only A LM self-training way | 41 | ||||
| 10篇softmax，CrossEntropyLoss替代方法论文合集 | 40 | ||||
| Cyclical Stochastic Gradient MCMC and snapshot ensemble | 39 | ||||
| Unsupervised Feature Learning via Non-Parametric Instance Discrimination | 38 | ||||
| An Image is Worth 16x16 Words Transformers for Image Recognition at Scale | 37 | ||||
| Training independent subnetworks for robust prediction | 36 | ||||
| Active Learning for CNNs: A Core-Set Approach | 35 | ||||
| SimCLR A Simple Framework for Contrastive Learning Visual Representation | 34 | ||||
| UNITER: UNiversal Image-TExt Representation Learning | 33 | ||||
| Image Synthesis with a Single (Robust) Classifier | 32 | ||||
| Set Transformer A Framework for Attention-based Permutation-Invariant NN | 31 | ||||
| Consistency Regularization in Semi-Supervised Learning | 30 | ||||
| Did the model understand the question | 29 | ||||
| Rethinking Feature Distribution for Loss Functions in Image Classification | 28 | ||||
| Bootstrap your own latent: A new way to self supervised learning | 27 | ||||
| Hybrid Discriminative-Generative Training via Contrastive Learning(EBMs) | 26 | ||||
| A Multimodal Translation-Based Approach for Knowledge Graph Representation (ACL 2018) | 25 | ||||
| Deep Bayesian Active Learning with Image Data (ICML 2017) The power of ensembles for active learning in image classification (CVPR 2018) |
24 | ||||
| SCAN Learnrnning to Classify Images without Labels (ECCV 2020) | 23 | ||||
| Unsupervised Question Answering by Cloze Translation (ACL 2019) | 22 | ||||
| Phrase-Based & Neural Unsupervised Machine Translation (EMNLP 2018) | 21 | ||||
| MixUp as Locally Linear Out-Of-Manifold Regularization (AAAI 2019) | 20 | ||||
| Manifold Mixup: Better Representations by Interpolating Hidden States ICML2019 | 19 | ||||
| Bag of Tricks for Image Classification with CNN | 18 | ||||
| On Mixup Training Improved Calibration for DNN | 17 | ||||
| BERT: Pre-training of Deep Bidirectional Transformers | 16 | ||||
| Rationalizing Neural Predictions (EMNLP2016) | 15 | ||||
| Attention is all you need, Transformer (NIPS 2017) | 14 | ||||
| Learn To Pay Attention (ICLR 2018) | 13 | ||||
| A Self-Training Method for MRC with Soft Evidence Extraction(ACL 2019) | 12 | ||||
| Deep Fool(CVPR2016) 和 Deep Defense(NIPS 2018) | 11 | ||||
| R-Trans RNN Transformer Network for 中文机器阅理解(IEEE-Access) | 10 | ||||
| 一系列Energy-based models 能量模型论文摘要简介 | 9 | ||||
| Implicit Generation and Modeling with EBM(NIPS 2019) | 8 | ||||
| MixMatch A Holistic Approach to Semi-supervised Learning(NIPS 2019) | 7 | ||||
| Obfuscated Gradients Give a False Sense of Security(ICML2017 best reward) | 6 | ||||
| Explaining and Harnessing Adversarial Examples(ICLR 2015) | 5 | ||||
| Imagenet-Trained CNNS are Biased Towards Texture(ICLR2018) | 4 | ||||
| Momentum Contrast for Unsupervised Visual Representation Learning(CVPR2020) | 3 | ||||
| Mixup: Beyond Empirical Risk Minimization(ICLR2018) | 2 | ||||
| Your Classifier is secretely an Energy Based Model(ICLR 2019) | 1（2020-08-19） |
代码
Paper List
可解释性
- Interpretable Convolutional Neural Networks
- Understanding Black-box Predictions via Influence Functions
Semi-Supervised Learning 半监督学习
- Regularization With Stochastic Transformations and Perturbations NIPS 2016
- Temporal Ensembling ICLR 2017
- Virtual Adversarial Training ICLR 2016
- Mean teachers are better role models: weight-averaged consistency targets NIPS 2017
- Realistic Evaluation of Deep Semi-SL NIPS 2018
- Deep co-training, ECCV 2018
- There Are Many Consistent Explanations Of Unlabeled Data why you should average ICLR 2019
- MixMatch A Holistic Approach to Semi-supervised Learning (NIPS 2019)
Softmax CrossEntropy Variants 交叉熵变种
- An Image is Worth 16x16 Words Transformers for Image Recognition at Scale
- Learn to Pay Attention (ICLR 2018)
- Large-Margin Softmax Loss for Convolutional Neural Networks ICML2016 https://arxiv.org/abs/1612.02295
- A Discriminative Feature Learning Approach for Deep Face Recognition ECCV 2016 https://link.springer.com/chapter/10.1007/978-3-319-46478-7_31
- Large Margin Deep Networks for Classification NIPS2018 https://arxiv.org/abs/1803.05598
- Rethinking Feature Distribution for Loss Functions in Image Classification CVPR 2018 http://arxiv.org/abs/1803.02988
- Max-Mahalanobis Linear Discriminant Analysis Networks http://arxiv.org/abs/1802.09308 ICML2018
- Rethinking Softmax Cross-Entropy Loss for Adversarial Robustness ICLR 2020 https://arxiv.org/pdf/1905.10626.pdf
- Redesigning the Classification Layer by Randomizing the Class Representation Vectors ICLR2021 under review https://openreview.net/forum?id=6_FjMpi_ebO
- Rethinking Feature Discrimination and Polymerization for Large-scale Recognition NIPS 2017 Deep Learning Workshop https://arxiv.org/abs/1710.00870
- Unsupervised Feature Learning via Non-Parametric Instance-level Discrimination http://arxiv.org/abs/1805.01978 CVPR 2018
- RBF-Softmax: Learning Deep Representative Prototypes with Radial Basis Function Softmax, ECCV 2020
Active Learning 主动学习
- Active Learning for CNNs: A Core-Set Approach
- Deep Bayesian Active Learning with Image Data (ICML 2017)
- The power of ensembles for active learning in image classification (CVPR 2018)
多模态
- UNITER: UNiversal Image-TExt Representation Learning
- A Multimodal Translation-Based Approach for Knowledge Graph Representation (ACL 2018)
Transfoormer
- Attention is All you need (NIPS 2017)
- BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
稳健性 Robustness
- Explaining and Harnessing Adversarial Examples (ICLR 2015)
- Deep Fool (CVPR2016)
- Deep Defense (NIPS 2018)
- Obfuscated Gradients Give a False Sense of Security (ICML2017 best reward)
- Adversarial Examples Are Not Bugs, They Are Features
- Image Synthesis with a Single (Robust) Classifier
- Adversarial Examples Improve Image Recognition
- You Only Propagate Once Accelerating AT via Maximal Principle
QA/MRC 机器阅读理解和问答
- R-Trans RNN Transformer Network for 中文机器阅读理解 (IEEE-Access)
- A Self-Training Method for MRC with Soft Evidence Extraction (ACL 2019)
- Did the model understand the question
- Rationalizing Neural Predictions (EMNLP2016)
图神经网络
- Graph Agreement Models for Semi-Supervised Learning
- Be More with Less: Hypergraph Attention Networks for Inductive 文本分类
- Text Level Graph Neural Network for Text Classification
- Graph Convolutional Networks for Text Classification
生成模型
GAN 生成式对抗网络
- BigGAN: Large Scale GAN Training for High Fidelity Natural Image Synthesis
- On the steerability of generative adversarial networks
自编码器
- Auto-Encoding Variational Bayes
分析
- CNN-Generated Images Are Surprisingly Easy to Spot.. For Now
神经网络稀疏化
- Learning sparse neural networks through L0 regularization
文本分类
- Be More with Less: Hypergraph Attention Networks for Inductive 文本分类
- Text Level Graph Neural Network for Text Classification
- Graph Convolutional Networks for Text Classification
无监督学习 Unsupervised Learning
- Unsupervised Question Answering by Cloze Translation (ACL 2019)
- Phrase-Based & Neural Unsupervised Machine Translation (EMNLP 2018)
- SCAN Learnrnning to Classify Images without Labels (ECCV 2020)
- Text Classification Using Label Names Only A LM self-training way
Contrastive Learning 对比学习
- Unsupervised Feature Learning via Non-Parametric Instance Discrimination
- Momentum Contrast for Unsupervised Visual Representation Learning (CVPR2020)
- SimCLR A Simple Framework for Contrastive Learning of Visual Representation
- Bootstrap your own latent: A new way to self supervised learning
- Hybrid Discriminative-Generative Training via Contrastive Learning(EBMs)
- What Makes for Good Views for Contrastive Learning
- Viewmaker Networks Learning Views for Unsupervised Representation Learning
EBM 能量模型
- Implicit Generation and Modeling with EBM (NIPS 2019)
- Your Classifier is secretely an Energy Based Model (ICLR 2019)
- Hybrid Discriminative-Generative Training via Contrastive Learning(EBMs)
Tricks 技巧
- Bag of Tricks for Image Classification with CNN
MixUp 数据融合
- Mixup: Beyond Empirical Risk Minimization (ICLR2018)
- MixUp as Locally Linear Out-Of-Manifold Regularization (AAAI 2019)
- Manifold Mixup: Better Representations by Interpolating Hidden States ICML2019
- On Mixup Training Improved Calibration
集成学习
- Cyclical Stochastic Gradient MCMC
- snapshot ensemble
- Training independent subnetworks for robust prediction
- Averaging Weights Leads to Wider Optima and Better Generalization Arxiv
其他
- Set Transformer: A Framework for Attention-based Permutation-Invariant NN
- Imagenet-Trained CNNS are Biased Towards Texture (ICLR2018)