There are many cases where we get the requirements of probabilistic models and techniques in neural networks. These requirements can be filled up by adding probability layers to the network that are provided by TensorFlow.