Diving into the trainer

As previously mentioned, for this linear regression application, we will be using the SDCA trainer. SDCA stands for Stochastic Dual Coordinate Ascent and if you may recall, we used the logistic regression version of this trainer in the example in Chapter 2, Setting Up the ML.NET Environment.

To the average reader, all four words that comprise SDCA might be unknown, so let's break down what each means to give better clarity to what happens when you utilize an SDCA trainer. Starting with Stochastic, which, in other words, means unpredictability. And in the case of machine learning, it means attempting to probabilistically predict the error function and feed random samples from your training set into the optimizer. The use of Dual Coordinate means two variables are coupled when training the model. As you have probably guessed, this makes the model much more complex but doesn't require any extra work to be utilized. Lastly, Ascent refers to maximizing the value of the error function.