ENH: random.multivariate_normal should broadcast input

Reproducing code example:

This works well:

m = np.random.normal(size=3)
x = np.random.normal(loc=m, scale=1)

This does not:

P = [[1,0.5],[0.5,1]]
M = np.random.multivariate_normal(mean=[0,0], cov=P, size=3)
X = np.random.multivariate_normal(mean=M, cov=P)

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-157-c99a31aba9db> in <module>
      1 M = np.random.multivariate_normal(mean=[0,0], cov=P, size=3)
----> 2 X = np.random.multivariate_normal(mean=M, cov=P)

mtrand.pyx in numpy.random.mtrand.RandomState.multivariate_normal()

ValueError: mean must be 1 dimensional

Numpy/Python version information:

1.17.2 / 3.7.4 (default, Aug 13 2019, 20:35:49) [GCC 7.3.0]

[eric-wieser]

I think I can phrase this as a feature request - you’d like multivariate_normal to work on stacks of matrices. So today, it takes mean of shape (N,) and cov of shape (N,N), and produces samples of shape size + (N,). The feature request would be to broadcast for mean of shape (..., N) and cov of shape (..., N, N), and produce samples of shape (..., N).

Does that sound reasonable?

---

Also, can you replace the traceback with text, so that it’s searchable?

[CloseChoice]

I would like to work on this