Skip to content
No description, website, or topics provided.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
Posttraining.py
README.md Update README.md May 9, 2019
SGDR.py
architecture1.png Add files via upload Apr 20, 2019
architecture2.png Add files via upload Apr 20, 2019
augmentation.py Add files via upload Apr 20, 2019
drop_block.py
mixup_generator.py Add files via upload Apr 20, 2019

README.md

EENA

The project is the post-training of the neural architectures designed by EENA (Efficient Evolution of Neural Architectures).
If you want to execute this project, you can run the Posttraining.py with the default configurations.
Serveral configurations:
>Standard augmentation: We normalize the images using channel means and standard deviations for preprocessing and apply a standard data augmentation scheme (zero-padding with 4 pixels on each side to obtain a 40*40 pixels image, then randomly cropping it to size 32*32 and randomly flipping the image horizontally).
>Cutout: n_holes = 1, length = 16.
>Mixup: α = 1.
The model is trained on the full training dataset until convergence using SGDR with a batch size of 128.
The final test error on CIFAR-10 can achieve around 2.56.

You can’t perform that action at this time.