Hands-On GPU Computing with Python

Training Fashion-MNIST for 100 epochs

To use 100 training cycles, we again to do a small change to the code:

model.fit(train_images, train_labels, epochs=100)

The same process will happen, but for 100 epochs, so it will take some time:

There's that sandal again! We see a prediction percentage improvement by 20%. Remember that, still, it is not the deciding factor. It is the red color that shows it's wrong. We humans can easily infer this. But there are far more complex objects that can be used instead of fashion objects, where even we humans require careful precision (we will explore that in the next chapter):

Here are multiple predictions ...