Reproduce the results in paper "Deep Residual Learning for Image Recognition", http://arxiv.org/abs/1512.03385 with the variants proposed in "Identity Mappings in Deep Residual Networks", https://arxiv.org/abs/1603.05027.
The train error shown here is a moving average of the error rate of each batch in training. The validation error here is computed on test set.
Download model: Cifar10 ResNet-110 (n=18)