The purpose of this project is to implement a deep neural classifier containing convolutional layers (as well as other layers). The dataset to be used is Fashion-MNIST. The Keras interface (to TensorFlow) must be used for this assignment. The ten classes in the dataset are listed at

https://keras.io/datasets/

For this assignment, you must provide

1)a display of some of the items in the datasets

2)three convolutional layers of different numbers of output units

3)two fully-connected layers, with appropriate dropouts in between

4)three different configurations of the model, using SGD, Adam, and Adagrad optimizers,

and corresponding training runs

5)a confusion matrix (predicted classes against actual (truth) classes) for the test data

6)conclusions, from the confusion matrix, as to which pairs of items the model confuses

(write these conclusions inside a markdown cell, at the end of the notebook).

Do in the jupyter notebook