Materials for transfer learning 中文版, English version
update:
- (2020,9,28) 新增DA paper
- (2020,9,6) 新增DA paper, 科研方法论相关视频
- (2020,8,29) 新增龙老师ccdm 2020报告视频
本部分内容适合初学者,将一些本领域中的经典论文按照时间线进行分类、梳理,分为浅层域适应、深度域适应、对抗域适应和域适应领域四部分。
针对每一部分,列举了3-4篇经典论文,建议详读这些经典论文,泛读这些经典论文的后续论文,并对其中的部分算法进行实现。
预期学习时间为2-3个月, 详细计划安排见入门参考
围绕这些论文,曾有一个相应的讨论班,相关的日程和资料如下:
适用深度网络的数据集
- mnist, svhn, digit, mnistm, cifar, stl (以上皆为.mat格式) 链接
- Office, Office-Home, VisDA-C, Office-Caltech from the official websites.
适用非深度网络的数据集(传统方法)
CCF推荐会议每年的举办时间会有稍稍的不同,此列表收集了当年的CCF推荐列表的截稿时间,包括了全部的CCF会议deadline和CCF期刊的special issue, 可作为一个近似参考,详细时间及内容建议查询官网确认。 链接:Call4Papars
- 龙明盛 清华大学
- 庄福振 中科院计算所
- 张宇 南方科技大学
- 李汶 ETH
- 王晋东 微软亚洲研究院
- 张磊 重庆大学
- Judy Hoffman Georgia Tech
- Kate Aaenko Boston University
- Sinno Jialin Pan NTU
- Kuniaki Saito Boston University(Ph.D)
- Zhao Han CMU
- 宫博庆 Google Research
- 督工 认知模型 链接
- 沈向洋 you are what you read 链接
- 沈向洋 how to read papers 7.18(私有), 文字版
- 王井东 how to read papers 7.21(私有, 密码同上)
- 袁路 how to read papers 7.24(私有,密码同上)
- 陈栋 how to read papers 7.27(私有,密码同上)
- 杨蛟龙 how to read papers 7.30(私有,密码同上)
- 胡瀚 how to read papers 8.2(私有,密码同上)
- 陈东东 how to read papers 8.5(私有,密码同上)
- 秦涛 do high-quality research (私有,密码同上)
- 杨强, 从 0 到 1,迁移学习如何登上今日高峰?链接
- [龙明盛 CCDM 2020] 视频 , ppt
- VALSE Webinar 20-19期 迁移学习 (个人非常推荐, 对新手不友好,对进阶有帮助,质量很高!) 视频, 报告简介
- 龙明盛_NJU2019 Transfer Learning Theories and Algorithms
- 龙明盛 Valse 2019 Transfer Learning_From Algorithms to Theories and Back
- 游凯超 智源论坛 2019 领域适配前沿研究--场景、方法与模型选择
- 王玫 2019 deep_domain_adaptation
- 吴恩达 NIPS 2016 Nuts and bolts of building AI applications using Deep Learning
Title | Conference/journel + year | Code | Keywords | Benenit for us |
---|---|---|---|---|
Neural transfer learning for natural language processing(paper) | 2019 PDH thesis | NLP, transfer lerning | very detailed related work | |
When Semi-Supervised Learning Meets Transfer Learning: Training Strategies, Models and Datasets(paper) | SSL, TL, experiments | many results related to multiple SSL methods can be seen in this paper | ||
Unsupervised Transfer Learning for Spatiotemporal Predictive Networks (paper) | ICML 2020 | |||
Estimating Generalization under Distribution Shifts via Domain-Invariant Representations (paper) | ICML 2020 | code | new theory | recommend to read |
Implicit Class-Conditioned Domain Alignment for Unsupervised Domain Adaptation (paper) | ICML 2020 | code | ideas from theory | recommend to read |
LEEP: A New Measure to Evaluate Transferability of Learned Representations (paper) | ICML 2020 | new metric for transferability | easy to use for other tasks | |
Label-Noise Robust Domain Adaptation | ICML2020 | the author is a rising star | ||
Progressive Graph Learning for Open-Set Domain Adaptation (paper) | ICML 2020 | code | open set DA | |
Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation (paper) | ICML 2020 | code | source-free DA | recommend to read, new trneds |
Graph Optimal Transport for Cross-Domain Alignment (paper) | ICML 2020 | graph for DA | connenction with GCN | |
Learning Deep Kernels for Non-Parametric Two-Sample Tests (paper) | ICML 2020 | code | extend MMD to deep | |
Adversarial-Learned Loss for Domain Adaptation | AAAI 2020 | noisy label, adversarial learning | ||
Transfer Learning for Anomaly Detection through Localized and Unsupervised Instance Selection | AAAI 2020 | transfer learning, anamaly detection | ||
Dynamic Instance Normalization for Arbitrary Style Transfer | AAAI 2020 | dynamic instance normalization | ||
AdaFilter: Adaptive Filter Fine-Tuning for Deep Transfer Learning | AAAI 2020 | gated output, fine-tune | ||
Bi-Directional Generation for Unsupervised Domain Adaptation | AAAI 2020 | differert feature extractor, different classifiers | connection with ICML 2019, the third term | |
Discriminative Adversarial Domain Adaptation | AAAI 2020 | discriminative information with adversarial learning | ||
Domain Generalization Using a Mixture of Multiple Latent Domains | AAAI 2020 | |||
Multi-Source Distilling Domain Adaptation | AAAI 2020 | multi-source | ||
Unsupervised Intra-domain Adaptation for Semantic Segmentation through Self-Supervision | CVPR 2020 | code | Entropy adversarial based | |
Rethinking Class-Balanced Methods for Long-Tailed Visual Recognition from a Domain Adaptation Perspective | CVPR 2020 | long-tailed | ||
Unsupervised Domain Adaptation via Structurally Regularized Deep Clustering | CVPR 2020 | code | cluster | |
Stochastic Classifiers for Unsupervised Domain Adaptation | CVPR 2020 | stochastic two classifiers | simialer to MCD | |
Progressive Adversarial Networks for Fine-Grained Domain Adaptation | CVPR 2020 | fine-grained | similar to mutil-aspect opinion analysis | |
Model Adaptation: Unsupervised Domain Adaptation without Source Data | CVPR 2020 | Recommend to read, new problems | ||
Towards Inheritable Models for Open-Set Domain Adaptation | CVPR 2020 | code | ||
Joint Disentangling and Adaptation for Cross-Domain Person Re-Identification | ECCV 2020 | |||
Extending and Analyzing Self-Supervised Learning Across Domains (paper) | ECCV 2020 | |||
Dual Mixup Regularized Learning for Adversarial Domain Adaptation (paper) | ECCV 2020 | |||
Label Propagation with Augmented Anchors: A Simple Semi-Supervised Learning baseline for Unsupervised Domain Adaptation (paper | ECCV 2020 | code | SSL reguralization, Anchors | new methods, good writings |
Do Adversarially Robust ImageNet Models Transfer Better? | arvix 2020 | code | Many experiments | |
Visualizing Transfer Learning | arvix 2020 | interesting | ||
A SURVEY ON DOMAIN ADAPTATION THEORY:LEARNING BOUNDS AND THEORETICAL GUARANTEES (paper) | arvix 2020 | theory | ||
SpotTune: Transfer Learning through Adaptive Fine-tuning (paper) | CVPR 2019 | code | dynamic routing is a general method | |
Parameter Transfer Unit for Deep Neural Networks (paper) | PAKDD 2019 best paper | good idea, recommened to read | ||
Heterogeneous Domain Adaptation via Soft Transfer Network (paper) | ACM MM 2019 | |||
Information-Theoretical Learning of Discriminative Clusters for Unsupervised Domain Adaptation (paper) | ICML 2012 | |||
Joint Disentangling and Adaptation for Cross-Domain Person Re-Identification (paper) | arvix 2020 | Good ideas | ||
Towards Recognizing Unseen Categories in Unseen Domains (paper) | arvix 2020 | new problems | ||
MiCo: Mixup Co-Training for Semi-Supervised Domain Adaptation (paper) | arvix 2020 | good framework | ||
Dynamic Knowledge Distillation for Black-box Hypothesis Transfer Learning (paper | arvix 2020 | |||
Simultaneous Semantic Alignment Network for Heterogeneous Domain Adaptation (paper) | ACM MM 2020 | code | ||
Learning from a Complementary-label Source Domain: Theory and Algorithms(paper) | arvix 2020 | code | novel idea | |
Class-Incremental Domain Adaptation(paper) | ECCV 2020 | new problems | ||
Class-incremental Learning via Deep Model Consolidation (paper) | WACV 2020 | |||
Adversarial Graph Representation Adaptation for Cross-Domain Facial Expression Recognition (paper) | ACM MM 2020 | similar idea with us | ||
A Review of Single-Source Deep Unsupervised Visual Domain Adaptation paper | arvix 2020 | Review | a good review! It contains many results of the state-of-the-art method |
Title | Conference/journel + year | Code | Keywords | Benenit for us |
---|---|---|---|---|
Collaborative Graph Convolutional Networks: Unsupervised Learning Meets Semi-Supervised Learning | AAAI 2020 | unsupervised learning, semi-supervised learning | ||
Self-supervised Label Augmentation via Input Transformations | ICML 2020 | code | self-supervised | ideas can be used to many tasks |
Learning with Multiple Complementary Labels (paper) | ICML 2020 | |||
Deep Divergence Learning (paper) | ICML 2020 | divergence | ||
Confidence-Aware Learning for Deep Neural Networks (paper) | ICML 2020 | code | confidence | |
Continual Learning in Human Activity Recognition:an Empirical Analysis of Regularization (paper) | ICML workshop | code | Continual learning bechmark | |
Automated Phrase Mining from Massive Text Corpora (paper) | ||||
Adversarially-Trained Deep Nets Transfer Better(paper | arvix 2020 | new findings | ||
Learning to Combine: Knowledge Aggregation for Multi-Source Domain Adaptation | arvix (paper) | same ideas with us |
Title | Conference + year | speaker | Benenit for us |
---|---|---|---|
Weakly Supervised Domain Adaptation with Deep Learning (link) | ACM MM 2016 | Xiaogang Wang |