Package 'rvfl'

Dropout Layer

Description

Applies dropout to the input data for regularization during training. The dropout rate determines the probability of retaining a unit.

Usage

dropout_layer(x, dropout = 0, seed = 42)

Arguments

x	A numeric matrix of input data.
dropout	A numeric value between 0 and 1 representing the dropout rate. Default is 0.
seed	An integer seed for random number generation. Default is 42.

Value

A matrix of the same shape as x with dropout applied.

Examples

# Apply dropout to a matrix
x <- matrix(runif(20), nrow = 5)
dropout_layer(x, dropout = 0.5)

---

Compute Fitted Values for RVFL2 Model

Description

Returns the fitted (predicted) values for the RVFL2 model on the given data.

Usage

## S3 method for class 'rvfl2'
fitted(object, x)

Arguments

object	An object of class rvfl2 (the fitted model).
x	A numeric matrix of input features (test data).

Value

A numeric vector of fitted values.

---

Simple and efficient implementation using the kernel trick on concatenated features

Description

Simple and efficient implementation using the kernel trick on concatenated features

Usage

krvfl(
  x,
  y,
  lambda = 0.1,
  nb_hidden = 100L,
  activation = c("relu", "sigmoid", "tanh", "linear"),
  sigma = 1,
  seed = NULL,
  ...
)

Arguments

x	Matrix of predictors (n x p)
y	Response vector (n x 1) or matrix (n x m)
lambda	Regularization parameter
nb_hidden	Number of hidden units
activation	Activation function
sigma	Scale parameter for weights
seed	Random seed
...	Additional arguments

Value

krvfl object

---

Predict method for krvfl objects

Description

Predict method for krvfl objects

Usage

## S3 method for class 'krvfl'
predict(object, newdata = NULL, ...)

Arguments

object	A krvfl object from krvfl()
newdata	Matrix of new predictors (m x p)
...	Additional arguments (unused)

Value

Predicted values (m x 1) or (m x q) matrix

---

Make Predictions with RVFL2 Model

Description

Predicts the target values using the fitted RVFL2 model on new input data.

Usage

## S3 method for class 'rvfl2'
predict(object, new_x, dropout = object$dropout, seed = object$seed)

Arguments

object	An object of class rvfl2 (the fitted model).
new_x	A numeric matrix of input features (test data).
dropout	A numeric value between 0 and 1 representing the dropout rate. Default is the value used during training.
seed	An integer seed for random number generation. Default is the value used during training.

Value

A numeric vector of predicted target values.

---

Print method for krvfl objects

Description

Print method for krvfl objects

Usage

## S3 method for class 'krvfl'
print(x, ...)

Arguments

x	A krvfl object
...	Additional arguments (unused)

---

Compute Residuals for RVFL2 Model

Description

Computes the residuals (observed - predicted) for the RVFL2 model on the given data.

Usage

## S3 method for class 'rvfl2'
residuals(object, x, y)

Arguments

object	An object of class rvfl2 (the fitted model).
x	A numeric matrix of input features (test data).
y	A numeric vector of observed target values (test labels).

Value

A numeric vector of residuals.

---

rvfl

Description

Generic function for RVFL models

Usage

rvfl(x, ...)

Arguments

x	Object input
...	Additional arguments passed to methods

---

Custom Regression Function (RVFL2)

Description

Fits a regression model using a Random Vector Functional Link (RVFL) network with dropout and regularization. The model includes both the original features and transformed features from a hidden layer.

Usage

rvfl2(
  x,
  y,
  nb_hidden = 50,
  sigma = 1,
  dropout = 0,
  seed = 42,
  lambda_1 = 0,
  lambda_2 = 0
)

Arguments

x	A numeric matrix of input features (training data).
y	A numeric vector of target values (training labels).
nb_hidden	An integer specifying the number of hidden units in the model. Default is 50.
sigma	A numeric value controlling the standard deviation of the initial random weights. Default is 1.
dropout	A numeric value between 0 and 1 representing the dropout rate. Default is 0.
seed	An integer seed for random number generation. Default is 42.
lambda_1	A numeric value controlling the L2 regularization on the input features. Default is 0.
lambda_2	A numeric value controlling the L2 regularization on the hidden layer features. Default is 0.

Value

A list of class rvfl2 containing the fitted model with the coefficients, hidden layer weights, and training parameters.

---

Summary method for krvfl objects

Description

Summary method for krvfl objects

Usage

## S3 method for class 'krvfl'
summary(object, ...)

Arguments

object	A krvfl object
...	Additional arguments (unused)

---

Summary of RVFL2 Model

Description

Displays a summary of the RVFL2 model, including the key hyperparameters.

Usage

## S3 method for class 'rvfl2'
summary(object)

Arguments

object

An object of class rvfl2 (the fitted model).

Value

None (prints summary information to the console).

---