TSVD Postprocessing of ASP

[cortner]

@tinatorabi -- all I did is put together an experimental postprocessing of the entire ASP path. It is not quite as fast as I had hoped, but it is ok-ish. Maybe a bit more profiling can make it better. If one were to store only every 10th iterate on the ASP path then I think the cost of ASP and TSVD postprocessing would be balanced. Maybe this is an option to introduce to the ASP solver?

I can't continue anymore tonight. Please feel free to push directly to this branch (you have write-access) to continue to integrate it with the ASP solver.

I suggest:

• introduce another solver parameter tsvd = true/false (default false for now)
• introduce another solver parameter nstore::Integer (default nstore = 100) to say how many ASP steps to save. You can sub-sample the full ASP path via somethime like every = len ÷ nstore and then [ path[1:every:end]; path[end]]

I hope this makes sense.

Also post here when you are done for the day so I know to pick this up again.

[tinatorabi]

I integrated tsvd with asp. I think this is all I need for the paper. :)