A machine learning model, neural netwrok or {[statistical algorithm}} will typically have a set of weights or parrameters that are fitted as part of traning. Howver, there are also often higehr-order parameters that determine the shape or architecture of the model. This may be simple such as the vlue of 'k' to use in the k-means algorithm, or more complex such as the number of layers, size of each layer and interconnection geometry in a {[deep neural network}}.
Used on page 172