Keras Deep Learning Cookbook

Plotting the training and testing results

First, let's plot the training results from the saved log file in the output folder. As can be seen, the loss comes down, but toward the end escalates to an abnormal value and again comes down.

MAE and mean_q steadily increases as a function of episodes.

1. mean_q is defined by the following function:

def mean_q(y_true, y_pred):  return K.mean(K.max(y_pred, axis=-1))

2. Let's now plot the reward and steps as a function of the episode:

In the preceding graph, it can be seen that the reward oscillates between ...