Fully connected neural network on MNIST dataset
a) Open the notebook fcn_MNIST.ipynb.
In this notebook we use a fully connected neural network to predict the handwritten digits of the MNIST dataset.
We have 4000 examples with 784 pixel values and 10 classes. Run the fist 3 cells.
b) Write the missing TensorFlow code in cell 4 for the first hidden layer.
c) Run the next two cells to store the graph and do a forward pass of the untrained network. Have a look at the network.
d) Now lets train the model. Finish the code to calculate the loss and accuracy for the validation set.