Concepts
--------

Precision and recal
'''''''''''''''''''

The main difference between precision and recall is that precision means of all the positives predicted, how many should really have been positive, while recall means how many of the real positives did you find.


Using functional keras api
''''''''''''''''''''''''''

Networks created by Sequential are easier to build and think about, but for more complex topologies, ones that may require more than one output or more than one input to be trained at the same time, the functional API is the recommended approach, since it allows you to create inputs as tensors and pass them to custom loss functions such as the MDN loss function that gives both the mean and the spread of the data given the dataset.