adaptive icon indicating copy to clipboard operation
adaptive copied to clipboard

Specify an API for defining the scale of point

Open basnijholt opened this issue 5 years ago • 1 comments

(original issue on GitLab)

opened by Jorn Hoofwijk (@Jorn) at 2018-11-19T10:33:48.135Z

Instead of having the loss function deal with scaling, we may pass the scaled points into the loss function, and specify some other mechanism to let the user define how to scale the points.

basnijholt avatar Dec 19 '18 16:12 basnijholt

We could start from at least unifying the treatment of scaling across learners.

For that we'd define a new class, say _Scale (or Scale, Scaler, etc), which would have the following interface:

On every tell we'd have something like:

if self._scale.update(new_point):
    self._recompute_all_losses

And then we'd pass self._scale.value to every computation of the loss function.

akhmerov avatar Mar 17 '19 21:03 akhmerov