simulate_regression_data: Simulate data for a Neural Network Structure.

In gmcmacran/NeuralNetworkSimulatoR: Creates Ideal Data for a Neural Network

Description

Simulate data for a Neural Network Structure.

Usage

simulate_regression_data(rows = 1000L, N = 5L, U = 5L, C = 0L,
  matrices, activations, noise = 0)

Arguments

rows	An integer scaler. The number of rows in simulated data.
N	An integer scaler. The number of normal random variables in X.
U	An integer scaler. The number of uniform random variables in X.
C	An integer scaler. The number of binary random variables in X.
matrices	A list. Each element is a matrix defining the structure the layer.
activations	A list. Each element is a activation function.
noise	A positive numeric scaler. The standard deviation of a normal random variable added to response. 0 is no noise.

Value

A matrix containing predictor variables and a response variable.

Examples

library(NeuralNetworkSimulatoR)

# Network with no hidden layers.
# Linear regression with weights 1, 2, and 3
M <- list(matrix(1:3, nrow = 3, ncol = 1))
A <- list(linear_R)
simData <- simulate_regression_data(
  rows = 1000L,
  N = 3L, U = 0L, C = 0L,
  matrices = M, activations = A,
  noise = 0
)
rm(A, M, simData)

# Network with 1 hidden layer.
# 10 nodes in first layer. Activation relu
# 5 nodes in second layer.  Activation elu_R (hidden layer)
M <- list(
  matrix(1:10, nrow = 10, ncol = 5),
  matrix(1:5, nrow = 5, ncol = 1)
)
A <- list(relu_R, elu_R)
simData <- simulate_regression_data(
  rows = 1000L,
  N = 5L, U = 5L, C = 0L,
  matrices = M, activations = A,
  noise = 0
)
rm(A, M, simData)

# Network with 2 hidden layers.
# 10 nodes in first layer. Activation relu
# 5 nodes in second layer.  Activation linear_R (hidden layer)
# 3 nodes in third layer.  Activation relu_R (hidden layer)
M <- list(
  matrix(1:10, nrow = 10, ncol = 5),
  matrix(1:5, nrow = 5, ncol = 3),
  matrix(1:3, nrow = 3, ncol = 1)
)
A <- list(relu_R, linear_R, relu_R)
simData <- simulate_regression_data(
  rows = 1000L,
  N = 5L, U = 5L, C = 0L,
  matrices = M, activations = A,
  noise = 0
)

rm(A, M, simData)

gmcmacran/NeuralNetworkSimulatoR documentation built on March 23, 2020, 12:51 a.m.