DOC: Draw more attention to which functions in random are convenience wrappers #13035
Conversation
This gives an error at least as far back as 1.12.x
d40cb74 to
277e7c4
Compare
| tuple as the first argument, use `numpy.random.standard_normal` instead. | ||
| If positive int_like arguments are provided, `randn` generates an array | ||
| of shape ``(d0, d1, ..., dn)``, filled | ||
| with random floats sampled from a univariate "normal" (Gaussian) |
There was a problem hiding this comment.
Don't need quotes around "normal".
There was a problem hiding this comment.
Inherited from the old docstring
| This is a convenience function. If you want an interface that takes a | ||
| tuple as the first argument, use `numpy.random.standard_normal` instead. | ||
| If positive int_like arguments are provided, `randn` generates an array | ||
| of shape ``(d0, d1, ..., dn)``, filled |
There was a problem hiding this comment.
Might as well justify the text a bit more.
There was a problem hiding this comment.
Makes the diff noisier and the two commits harder to separate, would prefer not to
numpy/random/mtrand/mtrand.pyx
Outdated
|
|
||
| Notes | ||
| ----- | ||
| For random samples from :math:`N(\\mu, \\sigma^2)`, use either of:: |
277e7c4 to
1e96346
Compare
|
Good suggestions, clearly I'm out of practice at writing docs. |
|
Changes can be viewed at matlab for numpy, rand, randn They seem to render nicely. |
|
I was hoping that zeros and ones would be links, not italic. |
|
Try to add |
|
@charris any more comments? The new material can be viewed at matlab for numpy, rand, randn, LGTM. |
|
@mattip: I'd like to fix those references, but haven't had a chance. Feel free to take over to do it, else I'll try in 10 hours or so. |
In response to twitter conversations that complain that `randn` is inconsistent with `zeros`, to which the answer is "that's because randn is a wrapper to make things look like matlab".
|
References updated |
1e96346 to
dac6fdc
Compare
|
The |
I've seen people complain in the past that
randn(M, N)is inconsistent withzeros((M, N)). They're right, but they're missing the point thatrandnexists to be consistent with Matlab, not to be self-consistent with numpy.Let's point those users more obviously to the function they're looking for.