Model Evaluation

In machine learning, it is common to define two distinct terms: parameter and hyper parameter. Parameters are properties that affect how a model makes predictions from data. Hyper parameters refer to how a model learns from data. Parameters can be learned from the data and modified dynamically. Hyper parameters are higher-level properties and are not typically learned from data. For a more detailed overview, refer to the book Python Machine Learning, by Sebastian Raschka and Vahid Mirjalili (Packt, 2017).

Problem Categories

Generally, there are two categories of problems solved by neural networks: classification and regression. Classification problems regard the prediction of the right categories from data; for instance, if the temperature is hot or cold. Regression problems are about the prediction of values in a continuous scalar; for instance, what the actual temperature value is?

Problems in these two categories are characterized by the following properties:

• Classification...