Resolve "(Learner1D) improve time complexity"

Closes #126 and #104

See merge request !139

See https://www.python.org/dev/peps/pep-0008/#imports

test all the different loss functions in each test

See merge request !135

and have fewer xfailing tests

Then equal losses will be sorted on x-coordinates.

make 'fname' a parameter to 'save' and 'load' only

Closes #122

See merge request !133

Joseph Weston authored 6 years ago and Bas Nijholt committed 6 years ago

This simplifies the API by making sure that the filenames are
only provided in one place (the calls to save and load).

Closes #122

* add 'curvature_loss_function' to the 'tutorial.custom_loss.rst'
* fix header styling
* fix doc-string

Resolve "(Learner1D) add possibility to use the direct neighbors in the loss"

Closes #119

See merge request !131

This abstracts the attribute 'nn_neighbors' away and makes it easier
for the user, because one can now just set the 'loss_per_interval'
and the 'nn_neighbors' will be set be default.

Because data is now in the 'BaseLearner'

This makes the code work with any number of neighbors. Now the
new triangle loss even works with nn_neigbors=0. I also added
that we pass the neighbors in all loss_per_interval functions.

Resolve "(LearnerND) allow any convex hull as domain"

Closes #114

See merge request !127

We were always dividing the total function execution time by
the number of cores, however sometimes the user sets 'ntasks'
so instead we should divide by '_get_max_tasks()'.

Also the future start_time wasn't measured in the case where
one uses a SequentialExecutor because the Future that is
returned already contains the result.