Question

First program i’m writing while reading various Haskell tutorials gives me a “Kind mis-match” error

import qualified Data.Vector as V

class SupervisedLearner l where
    learn :: l (Input n) -> Output n

data Input n = Supervised (V.Vector n ,V.Vector n) | Unsupervised (V.Vector n)
data Output n =  Regression n | KClass (V.Vector n) | Bernoulli (n, n)
newtype Perceptron  = Perceptron (V.Vector Float)

instance SupervisedLearner Perceptron  where
    learn = undefined

What’s confusing me is when i try to follow the type signature of the error,

Kind mis-match
The first argument of `SupervisedLearner' should have kind `*
                                                            -> *',
but `Perceptron' has kind `*'
In the instance declaration for `SupervisedLearner Perceptron'

I just can’t seem to understand where I should look to even begin to correct it.
So my question is two fold, where is the error, and a generally sense, am I using the Haskell typeclass system correctly?

Answer 1

According to your class definition, you need to define a method learn of type l (Input n) -> Output n to make l an instance of SupervisedLearner. So to make Perceptron an instance of SupervisedLearner, learn would have to have the type Perceptron (Input n) -> Output n.

But the Perceptron type doesn’t take any type arguments, so Perceptron (Input n) is not a valid type. That’s what the error message is complaining about (kind * means that a type doesn’t take any type arguments and kind * -> * means that a type takes one type argument).