Geometric Brownian Motion process seemingly can't be seeded / results can't be reproduced

[BenSchZA]

Hi there,

I created an MVP of a piece of code that has been causing issues, I don't think it's intended behaviour unless I'm misunderstanding some difference between GBM and BM processes. Whereas the BrownianMotion process will always return the same result when seeded, the GeometricBrownianMotion process does not. I have tried a couple of variations of the below code to test that it's not some bug in the order of code execution or seeding of the RNG. I did dig into the stochastic code a bit, but didn't find any obvious errors - maybe a maintainer of this codebase has a better idea?

MVP:

from stochastic import processes
import numpy as np

rng = np.random.default_rng(1)

gbm_process = processes.continuous.GeometricBrownianMotion(rng=rng)
gbm_sample = gbm_process.sample(1)

bm_process = processes.continuous.BrownianMotion(rng=rng)
bm_sample = bm_process.sample(1)

print(f"Geometric Brownian Motion sample: {gbm_sample}")
print(f"Brownian Motion sample: {bm_sample}")

Output:

Geometric Brownian Motion sample: [1.         0.62310484]
Brownian Motion sample: [0.         0.34558419]

Thanks!

[crflynn]

That should be fixed by https://github.com/crflynn/stochastic/pull/67 but I have yet to make a new release.

[BenSchZA]

Thanks for the quick response! Should have dug deeper into existing issues. I'll pull that down.