CODEWITHSUNDEEP
pythonnumpy
Demetri Pananos
Demetri Pananos

Reputation: 7404

How can I generate a gaussian process with correlation?

I can generate a gaussian process with np.random.normal(0,1). This GP is uncorrelated. How can I generate a gaussian process with correlation?

Upvotes: 0

Views: 1030

Answers (1)

Eric
Eric

Reputation: 97591

I think numpy.random.multivariate_normal(mu, cov) does just what you need.

You can also generate it with:

assert cov.shape == (N, N)
assert mu.shape == (N,)

L = np.linalg.cholesky(cov)
process = mu + L.T @ np.random.normal(0,1,N)

but this will fail for singular covariance matrices

Remember that you can only generate samples from a Gaussian process, since a process has infinitely many values

Upvotes: 1

Related Questions