Package 'rvmprior'

Title: Relevance Vector Machine with Flexible Spike-and-Slab Priors
Description: Fits Relevance Vector Machine (RVM) models for Normal, Epanechnikov, and Bickel prior distributions, and Gaussian or Mexican Hat kernels.
Authors: Alex Sousa [aut, cre], Rafael Oliveira [aut]
Maintainer: Alex Sousa <[email protected]>
License: MIT + file LICENSE
Version: 0.1.0
Built: 2026-07-22 10:54:33 UTC
Source: https://github.com/cran/rvmprior

Help Index

Fit plot of an rvm_prior object

Description

S3 visualization method for objects of class "rvm_prior". Displays the training data, the predicted curve, the true function (if available), and the selected relevance vectors.

Usage

## S3 method for class 'rvm_prior'
plot(x, f_true = NULL, ...)

Arguments

x

Object of class "rvm_prior".
f_true

Numeric vector or NULL. True function to overlay on the plot.
...

Additional arguments (ignored).

Value

Invisible ggplot object. The plot is printed as a side effect.

Examples

dados <- sinc_data(n = 100, SNR = 7)
fit <- rvm_prior(dados$x, dados$y, prior = "Epanechnikov",
                 kernel = "gaussian", f_true = dados$f_true)
plot(fit)

Fits an RVM model with flexible prior

Description

Main function of the rvmprior package. Fits a Relevance Vector Machine (RVM) model for regression with support for three types of priors (Normal, Epanechnikov, Bickel) and two kernel types (gaussian, mexican_hat).

Usage

rvm_prior(
  x,
  y,
  x_test = NULL,
  prior = c("Normal", "Epanechnikov", "Bickel"),
  kernel = c("gaussian", "mexican_hat"),
  kernel_param = NULL,
  prior_param = NULL,
  alpha = 0.95,
  show_plot = FALSE,
  f_true = NULL,
  ...
)

Arguments

x

Numeric vector. Predictor variable (unidimensional).
y

Numeric vector. Response variable. Must have the same length as x.
x_test

Numeric vector or NULL. Prediction points. If NULL, uses x.
prior

Character. Prior for the weights: "Normal", "Epanechnikov", or "Bickel".
kernel

Character. Kernel function: "gaussian" or "mexican_hat".
kernel_param

Numeric or NULL. Kernel parameter. If NULL, uses an automatic value.
prior_param

Numeric or NULL. Prior parameter. If NULL, uses an automatic value.
alpha

Numeric between 0 and 1. Spike probability. Default: 0.95.
show_plot

Logical. If TRUE, displays the fit plot. Default: FALSE.
f_true

Numeric vector or NULL. True function for comparison in the plot.
...

Additional arguments passed to rvm_spike_slab (iterations, tol).

Value

Invisible object of class "rvm_prior" with the fit results.

Examples

dados <- sinc_data(n = 100, SNR = 7)
fit <- rvm_prior(dados$x, dados$y, prior = "Epanechnikov", kernel = "gaussian")
print(fit)

Generate simulated data from the sinc function with Gaussian noise

Description

The sinc function is defined as sinc(x) = sin(x)/x, with limit sinc(0) = 1. Used in nonparametric regression simulations.

Usage

sinc_data(n = 100, from = -10, to = 10, SNR = NULL, sd_noise = 0.2, seed = 123)

Arguments

n

Integer. Number of points to generate. Default: 100.
from

Numeric. Lower bound of the x domain. Default: -10.
to

Numeric. Upper bound of the x domain. Default: 10.
SNR

Numeric or NULL. Desired signal-to-noise ratio. If NULL, uses sd_noise.
sd_noise

Numeric. Standard deviation of the noise. Used only when SNR is NULL. Default: 0.2.
seed

Integer. Seed for reproducibility. Default: 123.

Value

An object of class "sinc_data" with elements x, y, f_true, sd_noise and SNR.

Examples

dados <- sinc_data(n = 100, SNR = 7)
print(dados)