Neural Network training with Adam optimizer from scratch

Full code for training and testing of a simple neural network on the MNIST data set for recognition of single digits between 0 and 9 (Accuracy around 98 %). Everything is implemented from scratch, including the Adam optimizer. Make sure all the files are in your current folder and run "train.m".

Check out http://neuralnetworksanddeeplearning.com/index.html to learn about the theory of neural networks and https://arxiv.org/abs/1412.6980 to understand the Adam optimizer!