awesome-transfer-learning

Best transfer learning and domain adaptation resources (papers, tutorials, datasets, etc.)
https://github.com/eric-erki/awesome-transfer-learning

Last synced: 15 days ago
JSON representation

Surveys
Deep Transfer Learning
- Fine-tuning approach
  - Do Better ImageNet Models Transfer Better?
  - Using Pre-Training Can Improve Model Robustness and Uncertainty
- Feature extraction (embedding) approach
  - CNN Features off-the-shelf: an Astounding Baseline for Recognition
  - Taskonomy: Disentangling Task Transfer Learning
- Multi-task learning
  - Learning without forgetting
- Few-shot transfer learning
- Meta transfer learning
  - Transfer Learning via Learning to Transfer
- Applications
Unsupervised Domain Adaptation
Semi-supervised Domain Adaptation
- Subspace learning
  - Semi-supervised Domain Adaptation with Subspace Learning for Visual Recognition
Few-shot Supervised Domain Adaptation
- Adversarial methods
  - Few-Shot Adversarial Domain Adaptation
  - Augmented Cyclic Adversarial Learning for Domain Adaptation
- Embedding methods
  - Unified Deep Supervised Domain Adaptation and Generalization
Applied Domain Adaptation
- Physics
  - Identifying Quantum Phase Transitions with Adversarial Neural Networks
  - Automated discovery of characteristic features of phase transitions in many-body localization
- Audio Processing
Image-to-image
- Audio Processing
  - MNIST - M](https://drive.google.com/file/d/0B9Z4d7lAwbnTNDdNeFlERWRGNVk/view) vs [SVHN](http://ufldl.stanford.edu/housenumbers/) vs [Synth](https://drive.google.com/file/d/0B9Z4d7lAwbnTSVR1dEFSRUFxOUU/view) vs [USPS](http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multiclass.html#usps): digit images
  - CelebA
  - Office-Caltech dataset - 31 and Caltech-256 datasets. There are in total four domains: Amazon, Webcam, DSLR and Caltech.
  - Cityscapes dataset
  - CycleGAN datasets
  - pix2pix dataset
  - RaFD
  - DukeMTMC-reid - 1501](http://www.liangzheng.org/Project/project_reid.html): two pedestrian datasets collected at different places. The evaluation metric is based on open-set image retrieval.
Text-to-text
- Audio Processing
Digits transfer (unsupervised)
- Audio Processing
  - Open AI Retro Contest
  - Visual Domain Adaptation Challenge (VisDA)

Categories

Unsupervised Domain Adaptation 74 Surveys 44 Deep Transfer Learning 16 Applied Domain Adaptation 9 Image-to-image 8 Text-to-text 3 Few-shot Supervised Domain Adaptation 3 Digits transfer (unsupervised) 2 Semi-supervised Domain Adaptation 1

Sub Categories

Adversarial methods 45 Audio Processing 20 Optimal Transport 8 Kernel methods 7 Few-shot transfer learning 7 Subspace Learning 5 Embedding methods 3 Other 3 Applications 3 Theory 3 Fine-tuning approach 2 Feature extraction (embedding) approach 2 Physics 2 Autoencoder approach 2 Subspace learning 1 Self-Ensembling methods 1 Meta transfer learning 1 Multi-task learning 1