Machine Learning in Elixir

Learning with Optimization

Optimization is the search for the best. But how does this have anything to do with learning? Recall from Chapter 1, ​Make Machines That Learn​, that you can define a machine learning system as any computing system that’s capable of improving from experience with a specific task according to some arbitrary performance measure. The end goal is to have a system capable of providing predictions about unseen examples. Fundamentally, you want an interface that looks something like this:

def predict(input) do
  label = do_something(input)
  label
end

At its core, a trained machine learning system transforms inputs into labels. But what do these transformations look like? Is a machine learning algorithm synthesizing ...