Last active
December 21, 2015 10:04
-
-
Save sergeant-wizard/bc8577e2317cf4bdc192 to your computer and use it in GitHub Desktop.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
library(magrittr) | |
F <- function(y, theta) { | |
dnorm(y, mean = theta, sd = 1) | |
} | |
G0 <- function(theta) { | |
dnorm(theta, mean = 0, sd = 1) | |
} | |
sample_from_distribution <- function(min, max, distribution, threshold) { | |
while (TRUE) { | |
x <- runif(n = 1, min = min, max = max) | |
dice <- runif(n = 1, min = 0, max = 1) | |
if (dice < distribution(x) / threshold) { | |
return(x) | |
} | |
} | |
} | |
alpha <- 1 | |
num_iterations <- 4096 | |
min_theta <- -10 | |
max_theta <- 10 | |
y_array <- c( | |
rnorm(64, mean = -7.5, sd = 1), | |
rnorm(64, mean = 0, sd = 1), | |
rnorm(64, mean = 7.5, sd = 1) | |
) | |
# y_array <- rnorm(64, mean = 1, sd = 0.1) | |
num_samples <- length(y_array) | |
theta_array <- rep(0, num_samples) | |
theta_samples <- matrix(nrow = num_iterations, ncol = num_samples) | |
for (iteration in 1:num_iterations) { | |
for (y_index in 1:num_samples) { | |
y_i <- y_array[y_index] | |
# theta-wise array | |
q_over_b <- F(y_i, theta_array[-y_index]) | |
# constant | |
r_over_b <- 0.5 * alpha | |
b <- (sum(q_over_b) + r_over_b)^(-1) | |
r <- r_over_b * b | |
q <- q_over_b * b | |
dice <- runif(n = 1, min = 0, max = 1) | |
if (dice < r) { | |
# sample a new point from G0 * F | |
distribution_function <- function(theta) G0(theta) * F(y_i, theta) | |
# totally empirical | |
threshold <- max(distribution_function(c(0, 10))) * 3 | |
new_theta <- sample_from_distribution( | |
min = min_theta, | |
max = max_theta, | |
distribution = distribution_function, | |
threshold = threshold) | |
} else { | |
# sample from existing theta with probability proportonial to q | |
new_theta <- sample(size = 1, x = theta_array[-y_index], prob = q) | |
} | |
theta_array[y_index] <- new_theta | |
theta_samples[iteration, y_index] <- new_theta | |
} | |
} | |
null_axis <- rep(0, y_array %>% length) | |
plot(y_array, null_axis) | |
hist(y_array) | |
hist(theta_samples, probability = TRUE) | |
# histogram of unique elements | |
unique_elements <- apply(theta_samples, MARGIN = 1, FUN = function(x) x %>% unique %>% length) | |
hist(unique_elements) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment