Complete this project to gain a better understanding of how to add regularization to neural networks.
Using the code from the last lesson, add two more hidden layers, and add a dropout layer to each. How does that affect the training and validation accuracy that you can achieve?
Create a new dataset by taking the existing house dataset, and using data augmentation to add more rows. Does this get you better performace? How does it impact training time?
HINT:A good starting place would be to vary the inputs by 5% or so, and then experiment with keeping the sales price fixed; or also varying the sales price by x% in the same direction.