[MXNET-212] [R] Fix Random Samplers from Uniform and Gaussian distribution in R bindings

R-package/R/random.R

@@ -49,7 +49,7 @@ mx.runif <- function(shape, min=0, max=1, ctx=NULL) {
   if (!is.numeric(min)) stop("mx.rnorm only accept numeric min")
   if (!is.numeric(max)) stop("mx.rnorm only accept numeric max")
   ret <- mx.nd.internal.empty(shape, ctx)
-  return (mx.nd.internal.sample.uniform(min, max, shape=shape, out=ret))
+  return (mx.nd.internal.random.uniform(min, max, shape=shape, out=ret))
 }
 
 #' Generate nomal distribution with mean and sd.
@@ -73,5 +73,5 @@ mx.rnorm <- function(shape, mean=0, sd=1, ctx=NULL) {
   if (!is.numeric(mean)) stop("mx.rnorm only accept numeric mean")
   if (!is.numeric(sd)) stop("mx.rnorm only accept numeric sd")
   ret <- mx.nd.internal.empty(shape, ctx)
-  return (mx.nd.internal.sample.normal(mean, sd, shape=shape, out=ret))
+  return (mx.nd.internal.random.normal(mean, sd, shape=shape, out=ret))
 }

R-package/tests/testthat/test_random.R

@@ -0,0 +1,19 @@
+require(mxnet)
+
+context("random")
+
+test_that("mx.runif", {
+  X <- mx.runif(shape=50000, min=0, max=4, ctx=mx.ctx.default())
+  expect_equal(X>=0, mx.nd.ones(50000))
+  expect_equal(X<=4, mx.nd.ones(50000))
+  sample_mean = mean(as.array(X))
+  expect_equal(sample_mean, 2, tolerance=1e-2)
+})
+
+test_that("mx.rnorm", {
+  X <- mx.rnorm(shape=50000, mean=5, sd=1, ctx=mx.ctx.default())
+  sample_mean = mean(as.array(X))
+  sample_sd = sd(as.array(X))
+  expect_equal(sample_mean, 5, tolerance=1e-2)
+  expect_equal(sample_sd, 1, tolerance=1e-2)
+})