Mastering Predictive Analytics with scikit-learn and TensorFlow

Run the computational graph

To actually run the computational graph, first we initialize all of the variables in our program. The variables are the weights that are implicit in the fully_connected function. We run four epochs and for each epoch, we initialize our iterator object. We pass training x and training y, and then we run this loop. This loop will run as long as we have data in next_elementelement. So, we get the next 100 elements and then, in the next iteration, the next 100 elements, and so on. In every iteration, we run the training operation. Now, what this training operation does is ask the optimizer to adjust the parameters and the weights, a little bit in order to make better predictions.

The following screenshot shows the ...