Backend refactor - Githubissues

aneeshnaik commented 1 month ago

Major changes:

freesample is gone. The freesample case is now handled by a sequence of gridsample calls.
A lot of the grid logic (parsing cells and evaluating the PDF) is now handled in a new class (DensityGrid). The original LintSampler class creates one or several instances of DensityGrid, calls the 'evaluate' method on them, then can sample as normal.
In addition to the existing options for input cells, cells can be one or several pre-made DensityGrid instances, either with or without the densities already evaluated on them.
pdf is now an optional parameter in LintSampler. This is to allow the option of a pre-made DensityGrid which has already had densities evaluated. (error is raised if pdf is None and the DensityGrid instance doesn't have pdf evaluated).

aneeshnaik commented 1 month ago

For the most part I've written up the test suite and the docstrings.

michael-petersen commented 1 month ago

I'm now adjusting the examples in this pull request, to be consistent with the correct API.

aneeshnaik commented 1 month ago

[x] Finish 'Worked Example' page in docs/theory
[x] Add exceptions in DensityTree
[x] Add tests to cover DensityTree
[x] Implement DensityTree in DM notebook
[x] QMC example notebook
[x] Documentation for DensityTree (with examples) and base DensityStructure
[x] Fix DensityTree 1D functionality
[x] Construct densitytree with numbers (1D case), lists/tuples (kD case)
[ ] After finalising, go over DensityTree and DensityGrid attributes, check public/private
[x] Coverage -> 100%

aneeshnaik / lintsampler