Exercise

Defining the feature columns and input function

Last chapter, you created a DNN Classifer using the Banknote_Authentication dataset, in which you were able to fairly accurately assess whether a banknote was genuine or forged (response attribute: Class) based on several feature columns, including Var,Skew, Kurt, and Entropy.

The next task is to use a dropout technique to assess whether you can increase the accuracy of, and lower the chances of overfitting to, our DNN Classifier model using the TFEstimators API.

Instructions

100 XP

Define the four feature columns from the Banknote_Authentication dataset.
Create an input function, banknote_input_fn, by identifying the correct features and response variables.