hnn-0.1: A minimal Haskell Neural Network Library Contents Index

AI.HNN.Neuron

Contents Type Definitions, type class instances Neuron creation Transfer functions Neuron output computation Neuron learning with Widrow-Hoff (Delta rule)

Description

Neuron module, defining an artificial neuron type and the basical operations we can do on it

Synopsis

data Neuron = Neuron { threshold :: Double weights :: UArr Double func :: Double -> Double } createNeuronU :: Double -> UArr Double -> (Double -> Double) -> Neuron createNeuronHeavysideU :: Double -> UArr Double -> Neuron createNeuronSigmoidU :: Double -> UArr Double -> Neuron createNeuron :: Double -> [Double] -> (Double -> Double) -> Neuron createNeuronHeavyside :: Double -> [Double] -> Neuron createNeuronSigmoid :: Double -> [Double] -> Neuron heavyside :: Double -> Double sigmoid :: Double -> Double computeU :: Neuron -> UArr Double -> Double compute :: Neuron -> [Double] -> Double learnSampleU :: Double -> Neuron -> (UArr Double, Double) -> Neuron learnSample :: Double -> Neuron -> ([Double], Double) -> Neuron learnSamplesU :: Double -> Neuron -> [(UArr Double, Double)] -> Neuron learnSamples :: Double -> Neuron -> [([Double], Double)] -> Neuron

Type Definitions, type class instances

data Neuron

Our Artificial Neuron type Constructors Neuron threshold :: Double weights :: UArr Double func :: Double -> Double Instances Show Neuron

Neuron creation

createNeuronU :: Double -> UArr Double -> (Double -> Double) -> Neuron

Creates a Neuron with the given threshold, weights and transfer function

createNeuronHeavysideU :: Double -> UArr Double -> Neuron

Equivalent to `createNeuronU t ws heavyside'

createNeuronSigmoidU :: Double -> UArr Double -> Neuron

Equivalent to `createNeuronU t ws sigmoid'

createNeuron :: Double -> [Double] -> (Double -> Double) -> Neuron

Same as createNeuronU, with a list instead of an UArr for the weights (converted to UArr anyway)

createNeuronHeavyside :: Double -> [Double] -> Neuron

Same as createNeuronHeavysideU, with a list instead of an UArr for the weights (converted to UArr anyway)

createNeuronSigmoid :: Double -> [Double] -> Neuron

Same as createNeuronSigmoidU, with a list instead of an UArr for the weights (converted to UArr anyway)

Transfer functions

heavyside :: Double -> Double

The Heavyside function

sigmoid :: Double -> Double

The Sigmoid function

Neuron output computation

computeU :: Neuron -> UArr Double -> Double

Computes the output of a given Neuron for given inputs

compute :: Neuron -> [Double] -> Double

Computes the output of a given Neuron for given inputs

Neuron learning with Widrow-Hoff (Delta rule)

learnSampleU :: Double -> Neuron -> (UArr Double, Double) -> Neuron

Trains a neuron with the given sample, of the form (inputs, wanted_result) and the given learning ratio (alpha)

learnSample :: Double -> Neuron -> ([Double], Double) -> Neuron

learnSamplesU :: Double -> Neuron -> [(UArr Double, Double)] -> Neuron

Trains a neuron with the given samples and the given learning ratio (alpha)

learnSamples :: Double -> Neuron -> [([Double], Double)] -> Neuron

Trains a neuron with the given samples and the given learning ratio (alpha)

Produced by Haddock version 2.4.2