Problem 1.
(a) Use Tensorflow to construct a computation graph that performs training and predictionfor the MNIST dataset, with the following specifications:
(i) Single layer, 10 neurons, softmax + cross entropy objective function.
(ii) Three layers, [(50, ReLU), (50, ReLU), (10, Linear)], softmax + cross entropyobjective function. Use Stochastic Gradient Descent with minibatch size of 500.
(b) Use Keras (either plain Keras or tf.keras from Tensorflow 2.0) to redo the above twoexercises.