Synapses 7.0.0

Synapses

Neural network library in F#

Installation

Run dotnet add package Synapses --version 7.0.0 in the directory of your project.

Usage

Create a neural network

Open Synapses, call NeuralNetwork.init and provide the size of each layer:

open Synapses

let layers = [4; 6; 5; 3]

let neuralNetwork = NeuralNetwork.init(layers)

neuralNetwork has 4 layers. The first layer has 4 input nodes and the last layer has 3 output nodes. There are 2 hidden layers with 6 and 5 neurons respectively.

Get a prediction

let inputValues = [ 1.0; 0.5625; 0.511111; 0.47619 ]

let prediction = NeuralNetwork.prediction(neuralNetwork, inputValues)

prediction should be something like [ 0.829634; 0.699651; 0.454185 ]. Note that the lengths of inputValues and prediction equal to the sizes of input and output layers respectively.

Fit network

let learningRate = 0.5

let expectedOutput = [ 0.0; 1.0; 0.0 ]

let fitNetwork =
        NeuralNetwork.fit(neuralNetwork, learningRate, inputValues, expectedOutput)

fitNetwork is a new neural network trained with a single observation.

Save and load a neural network

let json = NeuralNetwork.toJson(fitNetwork)

Call NeuralNetwork.toJson on a neural network and get a string representation of it. Use it as you like. Save json in the file system or insert into a database table.

let loadedNetwork = NeuralNetwork.ofJson(json)

As the name suggests, NeuralNetwork.ofJson turns a json string into a neural network.

Customize a neural network

let activationF (layerIndex: int)
        : ActivationFunction =
        match layerIndex with
        | 0 -> ActivationFunction.sigmoid
        | 1 -> ActivationFunction.tanh
        | 2 -> ActivationFunction.leakyReLU
        | _ -> ActivationFunction.identity

let weightInitF (_layerIndex: int): float =
        System.Random().NextDouble()

let customizedNetwork =
        NeuralNetwork.customizedInit(layers, activationF, weightInitF)

The activation function of the neurons created with NeuralNetwork.init, is a sigmoid one. If you want to customize the activation functions and the weight distribution, call NeuralNetwork.customizedInit.

Encoding and decoding

One hot encoding is a process that turns discrete attributes into a list of 0.0 and 1.0. Minmax normalization scales continuous attributes into values between 0.0 and 1.0. You can use DataPreprocessor for datapoint encoding and decoding.

The first parameter of DataPreprocessor.init is a list of tuples (attributeName, discreteOrNot).

let setosaDatapoint =
        Map.ofList
            [ ("petal_length", "1.5")
              ("petal_width", "0.1")
              ("sepal_length", "4.9")
              ("sepal_width", "3.1")
              ("species", "setosa") ]

let versicolorDatapoint =
        Map.ofList
            [ ("petal_length", "3.8")
              ("petal_width", "1.1")
              ("sepal_length", "5.5")
              ("sepal_width", "2.4")
              ("species", "versicolor") ]

let virginicaDatapoint =
        Map.ofList
            [ ("petal_length", "6.0")
              ("petal_width", "2.2")
              ("sepal_length", "5.0")
              ("sepal_width", "1.5")
              ("species", "virginica") ]

let dataset = Seq.ofList
                [ setosaDatapoint
                  versicolorDatapoint
                  virginicaDatapoint ]
                
let dataPreprocessor = DataPreprocessor.init(
                            [ ("sepal_length", false)
                              ("sepal_width", false)
                              ("petal_length", false)
                              ("petal_width", false)
                              ("species", true) ],
                            dataset
                       )

let encodedDatapoints =
        Seq.map (fun datapoint -> DataPreprocessor.encodedDatapoint(dataPreprocessor, datapoint))
                dataset

encodedDatapoints equals to

[ [ 0.0     ; 1.0     ; 0.0     ; 0.0     ; 0.0; 0.0; 1.0 ]
  [ 1.0     ; 0.562500; 0.511111; 0.476190; 0.0; 1.0; 0.0 ]
  [ 0.166667; 0.0     ; 1.0     ; 1.0     ; 1.0; 0.0; 0.0 ] ]

Save and load the preprocessor by calling DataPreprocessor.toJson and DataPreprocessor.ofJson.