The author of HLearn is actually more interested in discovering a novel way of encoding machine learning problems using new type system features, as opposed to

The author of HLearn is actually more interested in discovering a novel way of encoding machine learning problems using new type system features, as opposed to choosing a more familiar encoding that exists in other languages, that Haskell would otherwise have no problems with. This locates the constraint on the authors ambitions relative to the desired abstraction level, rather than the performance level available