Unify behaviour of '''hessian''' for univariate and multivariate functions.

[Emilius12]

When working on this pull request, I encountered an error while attempting to compute the Hessian matrix of a univariate function. Upon investigation, I found that this error was introduced in this commit. I questioned whether it was advisable to retain this error. Modifying the behavior of this function seemed straightforward, but lacking in-depth knowledge of the potential ramifications of such a change, I sought the community's input.

[oscarbenjamin]

Can you explain what the problem is that is caused by that commit?

[Emilius12]

The issue is that the function '''hessian''' cannot work for univariate functions. My current solution for implementing convexity check for multivariate function is to discriminate cases but it would be better to have a unique behaviour

[oscarbenjamin]

It seems to work for univariate functions:

In [8]: hessian(f(x), [x])
Out[8]:
⎡ 2       ⎤
⎢d        ⎥
⎢───(f(x))⎥
⎢  2      ⎥
⎣dx       ⎦

In [9]: hessian(f(x, y), [x, y])
Out[9]:
⎡  2               2           ⎤
⎢ ∂               ∂            ⎥
⎢ ───(f(x, y))   ─────(f(x, y))⎥
⎢   2            ∂y ∂x         ⎥
⎢ ∂x                           ⎥
⎢                              ⎥
⎢  2               2           ⎥
⎢ ∂               ∂            ⎥
⎢─────(f(x, y))   ───(f(x, y)) ⎥
⎢∂y ∂x              2          ⎥
⎣                 ∂y           ⎦

[Emilius12]

Okay, I didn't realize that you had to call hessian(f, [x]) instead of hessian(f, x). Thanks for pointing that out! I think I'll be able to implement convexity checks for both univariate and multivariate functions now. Sorry for any inconvenience caused.