concept novel class in category Keras

This is an excerpt from Manning's book Probabilistic Deep Learning: With Python, Keras and TensorFlow Probability.
But in reality this is not always the case. Imagine that you provide the classifier with an image that shows neither Chantal nor Sara. The classifier has no other choice than assigning probabilities to the classes Chantal or Sara. But you would hope that the classifier shows its uncertainty by assigning more or less equal probabilities to the two possible but wrong classes. Unfortunately, this is often not the case when working with probabilistic NN models. Instead, often still quite high probabilities are assigned to one of the possible but wrong classes (see figure 1.8). To tackle this problem, in part 3 of our book we extend the probabilistic models by taking a Bayesian approach that can add an additional uncertainty that you can use to detect novel classes.
In this chapter, you’ll learn some approximation methods that allow fitting Bayesian variants of probabilistic DL models. This gives you a tool to detect if the predicted outcome distributions are uncertain. Bayesian DL models have the advantage that they can detect novel situations by expressing a larger epistemic uncertainty, which leads to a larger outcome uncertainty. You’ll see that Bayesian DL regression models report greater uncertainties in extrapolation regimes, and Bayesian DL classification models raise the uncertainty flag for novel classes, indicating that their predictions aren’t reliable.