densratio (version 0.2.1)

densratio: Estimate Density Ratio p(x)/q(x)

Description

Estimate Density Ratio p(x)/q(x)

Usage

densratio(x1, x2, method = c("uLSIF", "RuLSIF", "KLIEP"),
  sigma = "auto", lambda = "auto", alpha = 0.1, kernel_num = 100,
  fold = 5, verbose = TRUE)

Arguments

x1

numeric vector or matrix. Data from a numerator distribution p(x).

x2

numeric vector or matrix. Data from a denominator distribution q(x).

method

"uLSIF" (default), "RuLSIF", or "KLIEP".

sigma

positive numeric vector. Search range of Gaussian kernel bandwidth.

lambda

positive numeric vector. Search range of regularization parameter for uLSIF and RuLSIF.

alpha

numeric in [0, 1]. Relative parameter for RuLSIF. Default 0.1.

kernel_num

positive integer. Number of kernels.

fold

positive integer. Numer of the folds of cross validation for KLIEP.

verbose

logical (default TRUE).

Value

densratio object that contains a function to compute estimated density ratio.

Examples

# NOT RUN {
x1 <- rnorm(200, mean = 1, sd = 1/8)
x2 <- rnorm(200, mean = 1, sd = 1/2)

densratio_obj <- densratio(x1, x2)

new_x <- seq(0, 2, by = 0.05)
estimated_density_ratio <- densratio_obj$compute_density_ratio(new_x)

plot(new_x, estimated_density_ratio, pch=19)

# }