Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.
And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.
Stochastic behaviour even after setting random.seed()
See original GitHub issuenumpy.random.multivariate_normal() result is not reproducible even if numpy.random.seed() is set.
np.random.seed(9) class1 = np.random.multivariate_normal(mean, cov, n_samples) class2 = np.random.multivariate_normal(mean, cov, n_samples)
class1 won’t be the same as class2. When I generate mean and cov with np.random.rand() it is reproducible. cov is a positive semi-definite matrix I generated using np.dot(A,A.T) where A is generated with random.rand()
This was on Python 2.7.12 with Numpy 1.11.1 on Ubuntu 16.04.1
Issue Analytics
- State:
- Created 7 years ago
- Comments:5 (4 by maintainers)
Top Results From Across the Web
Top Related Medium Post
Top Related StackOverflow Question
Troubleshoot Live Code
Top Related Reddit Thread
Top Related Hackernoon Post
Top Related Tweet
Top Related Dev.to Post
Top Related Hashnode Post
OK, I think I will close this (though feel free to keep discussing). If you think a note in the documentation would have helped you, please do have a look and open a pull request to improve it!
The data generated from each run of the code is same but from calling the function multiple times they are not. @MSeifert04 I now tried to re-set the seed and it helps a lot. Although not identical the data generated is much closer in class1 and class2. @seberg I am guessing these small changes are from the stochastic nature of the linalg part.
Thank you for the quick and helpful answers.