Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/sungyubkim/GBML
A collection of Gradient-Based Meta-Learning Algorithms with pytorch
https://github.com/sungyubkim/GBML
cavia few-shot-learning gradient-based-meta-learning implicit-maml maml meta-learning meta-learning-algorithms neumann-approximation pytorch reptile
Last synced: 2 days ago
JSON representation
A collection of Gradient-Based Meta-Learning Algorithms with pytorch
- Host: GitHub
- URL: https://github.com/sungyubkim/GBML
- Owner: sungyubkim
- License: mit
- Created: 2019-10-15T14:38:55.000Z (about 5 years ago)
- Default Branch: master
- Last Pushed: 2019-12-09T07:20:34.000Z (almost 5 years ago)
- Last Synced: 2024-08-04T03:09:50.590Z (4 months ago)
- Topics: cavia, few-shot-learning, gradient-based-meta-learning, implicit-maml, maml, meta-learning, meta-learning-algorithms, neumann-approximation, pytorch, reptile
- Language: Python
- Homepage:
- Size: 1.41 MB
- Stars: 60
- Watchers: 3
- Forks: 8
- Open Issues: 5
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# GBML
A collection of Gradient-Based Meta-Learning Algorithms with pytorch
* [MAML](http://proceedings.mlr.press/v70/finn17a)
```python
python3 main.py --alg=MAML
```
* [Reptile](https://openai.com/blog/reptile/)
```python
python3 main.py --alg=Reptile
```
* [CAVIA](http://proceedings.mlr.press/v97/zintgraf19a)
```python
python3 main.py --alg=CAVIA
```
## Results on miniImagenet
* Without pre-trained encoder (Use 64 channels by default. The exceptions are in parentheses)
| | 5way 1shot | 5way 1shot (ours) | 5way 5shot | 5way 5shot (ours) |
| -------------- | ------------------- | ----------------- | ------------------- | ----------------- |
| MAML | 48.70 ± 1.84% | 49.00 % | 63.11 ± 0.92% | 65.18 % |
| Reptile | 47.07 ± 0.26% | 43.40 % | 62.74 ± 0.37% | - |
| CAVIA | 49.84 ± 0.68% (128) | 50.07 % (64) | 64.63 ± 0.53% (128) | 64.21 % (64) |
| iMAML | 49.30 ± 1.88% | - | - | - |
| Meta-Curvature | 55.73 ± 0.94% (128) | - | 70.30 ± 0.72% (128) | - |
* With pre-trained encoder (To be implemented.)
| | 5way 1shot | 5way 1shot (ours) | 5way 5shot | 5way 5shot (ours) |
| -------------- | ------------- | ----------------- | ------------- | ----------------- |
| Meta-SGD | 56.58 ± 0.21% | - | 68.84 ± 0.19% | - |
| LEO | 61.76 ± 0.08% | - | 77.59 ± 0.12% | - |
| Meta-Curvature | 61.85 ± 0.10% | - | 77.02 ± 0.11% | - |
## Dependencies
* Python >= 3.6
* Pytorch >= 1.2
* [Higher](https://github.com/facebookresearch/higher)
* [Torchmeta](https://github.com/tristandeleu/pytorch-meta)
## To do
* Add ~~ResNet~~ and Pre-trained encoder
* Add iMAML, Meta-Curvature