Diving deeper into Matlab’s endless built-in functions, I discovered (i.e. read) Mathworks‘ nnet manual. I usually abhor user manuals for specific programming languages, but Mathworks has made it an enjoyable read.

It is well-stocked with examples for specific functions and also cites references to algorithms. It generally seems to be suitable to be cited somewhere. Of course, it should only be sufficient as one of the starting points for further research. For example, lately I have been thinking about when to stop training the neural network on a specific data set. In their manual, it says (above pdf file, Page 5-53, Backpropagation Section, „Improving Generalization“):

Note that if the number of parameters in the network is much smaller than the
total number of points in the training set, then there is little or no chance of
overfitting. If you can easily collect more data and increase the size of the
training set, then there is no need to worry about the following techniques to
prevent overfitting. The rest of this section only applies to those situations in
which you want to make the most of a limited supply of data.

For the current data sets that I’m working with this means that I won’t have to care about overfitting with the agriculture data (900 and more records) but that I’ll have to take care when working with the sports science data (~50 records). I might even try some of the built-in techniques to counter overfitting.