Using Wide Residual Networks to get state-of-the-art results in CIFAR-10 dataset
Do not put too much attention on file names, I have been experimenting with different settings, so file names do not make sense.
Something is still not right, I cannot achieve accuracy as shown in the paper (https://arxiv.org/pdf/1605.07146.pdf), pytorch code of this repo: https://github.com/szagoruyko/wide-residual-networks/tree/master/pytorch
My keras model is build based on this repo: https://github.com/titu1994/Wide-Residual-Networks/
Graph of the current model (graph is missing some epochs from the beginning):
Download link for the trained model with best accuracy(around 200 epochs): https://drive.google.com/open?id=0BwpxSJhEGdqfV1M3NUZicXllZmc