Stickbreaking transform

Questions
1

I am trying to understand the algorithms used in Stickbreaking(), so I tried extracting the code for StickBreaking.forward_val to experiment with it. I can’t understand why it gives a different result when I do this, as in the following:

import numpy as np
import pymc3
from pymc3.distributions.transforms import StickBreaking

stick = StickBreaking()
print(pymc3.__version__)

def forward_py(x_, point=None):
    x = x_.T
    n = x.shape[0]
    lx = np.log(x)
    shift = np.sum(lx, 0, keepdims=True) / n
    y = lx[:-1] - shift
    return y.T

x = np.array([[0.31604495, 0.10538385, 0.00798379, 0.37937219, 0.19121521],
 [0.67614711, 0.14670955, 0.08485597, 0.07346115, 0.01882622]])
print(forward_py(x))
print(stick.forward_val(x))
2

I ran it locally and got the same result - what is the differences when you ran it?

3

Hi

These are the results I got. Completely different:

3.9.3
[[ 1.01931957 -0.07895548 -2.65915164 1.20195288]
[ 1.88084506 0.35288917 -0.19461027 -0.33880892]]
[[ 0.61428663 -0.60433991 -3.57610594 0.68511821]
[ 2.12241566 0.91010784 0.60919523 1.36150612]]

4

Hi Jungpen

Which of my results is the same as yours? Thanks.