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 2.26.