MNISTFullyConnectedTensorflowExample

This repo contains the code for the Second meetup of WLML. In this example we use a fully connected NN to solve the MNIST handwritten numbers classification task.

To view results run tensorboard --logdir ./tmp on terminal and follow instructions.

Files

Dataset

A utility that imports the MNIST data from remote and pre-processes for use in the tensorflow mode. In this module defines

• the size of the training image set train_size
• the size of the testing image set test_size
• the size of the images 'image_size'
• feed_dict_gen(BATCH_SIZE[integer], STEP[integer], test[any]): Gives the next batch according to the number of steps. If no value is given for test it will use the data from the training set else it will use the data from the test set

utils.py

defines few internal utilities

trainer.py

Defines the loss, step, optimizer and accuracy subgraphs.

model.py

Defines the model graph

main.py

Initializes TF graph, runs training and testing session.