McCullochPitts-class {deep}

The McCullochPitts neuron class, that implements the logic of the McCullochPitts neuron model.

Description

The McCullochPitts neuron class, that implements the logic of the McCullochPitts neuron model.

Arguments

inputs

The actual data to be fed to the nuron, this input's dimentions vary with the chosen weights dimentions.

ins

The list of vectors of inputs to the first layer in the network

outs

The list of vectors of outputs of the last layer in the network

epochs

How many rounds of training to run

tax

This is the learning rate, aka eta

maxErr

A contition to early stop the training process

Value

The computed value using the McCullochPitts model.

Vector of computed values of the same size of the last layer

Fields

ws

The matrix of weights that multiply the input vector, it can be a vector, a matrix or an array.

bias

The bias value.

Examples

# Create a dataset
dataset <- iris
dataset$Petal.Length <- NULL
dataset$Petal.Width <- NULL
dataset <- dataset[dataset$Species != "versicolor",]
dataset$Code <- as.integer(dataset$Species == "virginica")
dataset <- dataset[sample(20),]

# Create the neuron
neuron <- mcCullochPitts(c(1,1), 1)

# Train the neuron, takes a while
neuron$train(dataset[,c(1,2)], dataset[,'Code', drop=FALSE], epochs = 10)

# Check the output
neuron$output(c(1,2))

# See accuracy
dataset$Calc <- sapply(1:nrow(dataset), function(x) {
    as.integer(neuron$output(dataset[x,c(1,2)]))
})
length(which(dataset$Code==dataset$Calc))/nrow(dataset)