Add gradient to linear interpolant #68
Conversation
nevis/_interpolation.py
Outdated
| @@ -67,6 +67,27 @@ def __call__(self, x, y): | |||
| # Final result | |||
| return np.where(f1 == f2, f1, (y2 - y) * f1 + (y - y1) * f2) | |||
|
|
|||
| def grad(self, x, y): | |||
There was a problem hiding this comment.
Should we make it so that this returns the value and the gradient? I can't think of many scenarios where you just want the gradient, and the functions overlap quite a bit
There was a problem hiding this comment.
Could you also explain the returned format in the docstring? Maybe give an example too?
There was a problem hiding this comment.
I see... So is it better (and what you originally meant) to incorporate grad as an optional parameter of the function, which makes the function return extra fields when set True?
There was a problem hiding this comment.
You mean like f(x, y, True) would give you gradients?
Yeah we could do that!
Or I suppose we could write an entirely separate function (or callable class) ?
Up to you!
There was a problem hiding this comment.
In PINTS we have a __call__ for just f, and then a second function for f+df/dx: https://github.com/pints-team/pints/blob/master/pints/_error_measures.py#L12-L42
There's probably no right or wrong, just pick whatever you think is easiest!
There was a problem hiding this comment.
What about just outside, e.g. (-1, 0) ?
There was a problem hiding this comment.
BTW, do we need to make the function value a float?
Not sure what you mean? Python 3 does this automatically if needed
There was a problem hiding this comment.
Like -17.86381531 instead of array(-17.86381531). I encountered an error causes by this type issue when using NLopt (which is not Python implemented).
There was a problem hiding this comment.
What about just outside, e.g. (-1, 0) ?
Same result :-)
There was a problem hiding this comment.
Aha! Hadn't noticed that! Yeah make it a float
#27