Code to predict MNIST digits by implementing a Multi Layer Perceptron with a single hidden layer, without using any libraries like Tensorflow, Theano etc. Only linear algebra libraries like numpy have been used for the implementation.
The dataset has been downloaded from http://yann.lecun.com/exdb/mnist/ (four files). The 4 files should be extracted into a folder named data just outside the folder containing the main.py file i.e. the code in the file main.py reads the input data files from the folder '../data'. The train function trains the neural network given the training examples and saves the weights in a folder named weights in the same folder as main.py. The test function reads the saved weights and given the test examples, it returns the predicted labels.