[Bug]: maxwell

[dash96]

Bug Explanation

I recently learnt that, to verify if the random function is working alright one needs to also plot the histogram graph of the particular function along side the histogram graph of a back-end function for a large number of outputs to verify if they look similar.

I had pushed a Maxwell function that passed all the tests. but when I plotted the histogram graph as explained above they looked very different. The graph is as shown bellow,

The code for the same is mentioned in the next section.

Steps to Reproduce Bug

`%matplotlib inline

import matplotlib
import ivy
import jax
from ivy.functional.frontends.jax.random import _get_seed

current function in Ivy repository

def maxwell_current(key, shape=None, dtype="float64"):
    seed = _get_seed(key)
    # generate uniform random numbers between 0 and 1
    z = ivy.random_uniform(seed=seed, shape=shape, dtype=dtype)
    # applying inverse transform sampling
    x = (z**2) * ivy.exp(-(z**2) / 2)
    return x

edited version

def maxwell(key, shape, dtype= "float64"):
    seed = _get_seed(key)
    shape = shape + (3,)
    random_normal = ivy.random_normal(seed=seed, shape=shape, dtype=dtype)
    return ivy.vector_norm(random_normal, axis=-1)

n = 1000000
import matplotlib.pyplot as plt
x = maxwell_current((123,123), (n,))
y = maxwell((123,123), (n,))
z = jax.random.maxwell(jax.numpy.array([123,123], dtype='uint32'), shape=(n,))

num_bins = int(ivy.sqrt(n))
n0, bins0, patches0 = plt.hist(x, bins=num_bins, histtype='step', color='blue')
n, bins, patches = plt.hist(y, bins=num_bins, histtype='step', color='black')
n, bins, patches = plt.hist(z, bins=num_bins, histtype='step', color='red')

plt.plot()

Environment

In general, in colab.

Ivy Version

v1.1.9

Backend

• NumPy
• TensorFlow
• PyTorch
• JAX

Device

No response