Building our first DFN

So far, we have learned how a DFN works, and about the architecture, and aspects involved in training the network. In this section, we shall build our first DFN using TensorFlow. Building any deep learning would more or less involve the following steps:

1. Reading the input data and expected output data (labels)
2. Preparing the data in the required format (preprocessing)
3. Splitting the data into a training, validation, and testing set (a validation set is sometimes optional)
4. Building the model architecture graph along with the loss function and optimizer to update weights
5. Running a TensorFlow Session to iterate over data and train the network
6. Testing the accuracy of the model over test data

MNIST fashion data

...