Comparing-MNIST-Images

I wrote this to compare MNIST images generated by various Generative Adversarial Network's against the Original MNIST Data set after Martin Andrews suggested using a Mean Square Error loss to find the closest images in the data set to the image we want to check. One interesting thing it does show is that some images in the MNIST Test set are very similar to those in the Train set.