New Optimizer code#125

Merged

pavelkomarov merged 33 commits into

hack-the-optimizer-more

Jul 15, 2025

Commits on Jul 7, 2025

big changes! Basically fully rewrote the optimizer code. Getting same answers for spectraldiff, going to follow on with moves of some of the other modules, deleting now-unnecessary code as I go.
pavelkomarov
committed

Commits on Jul 8, 2025

moved rest of linear_model optimization
pavelkomarov
committed
moved finite difference to new optimizer
pavelkomarov
committed
moved smooth fd to new optimization
pavelkomarov
committed
moved TVR to new optimizer
pavelkomarov
committed
moved kalman module, to complete all 5
pavelkomarov
committed
addressed #82
pavelkomarov
committed
changed init_conds to search_space because it's a flashier, more memorable, descriptive variable name
pavelkomarov
committed
removed a couple unnecessary imports, reran basic_tutorial notebook (RMS numbers worsened because no longer excluding the beginning and ends of the vectors in calculation), and changed the evaluati…
pavelkomarov
committed
Merge pull request #126 from florisvb/notebooks-update-for-optimizer
pavelkomarov
authored
Merge branch 'master' of github.com:florisvb/PyNumDiff into notebooks-update-for-optimizer
pavelkomarov
committed
Merge branch 'master' of github.com:florisvb/PyNumDiff into hack-the-optimizer-more
pavelkomarov
committed
realized TV is defined as normalizing by m, not m-1 in the paper, so changed the code to agree
pavelkomarov
committed
Merge branch 'hack-the-optimizer-more' of github.com:florisvb/PyNumDiff into notebooks-update-for-optimizer
pavelkomarov
committed
fully updated notebook 2a
pavelkomarov
committed
notebook 2b now uses the new optimizer
pavelkomarov
committed
added search_space arg to test_polydiff in optimizer tests to lock down step_size, because when I expanded the search space to include step size, the answer changed
pavelkomarov
committed
Merge pull request #128 from florisvb/notebooks-update-for-optimizer
pavelkomarov
authored
importing the optimize function to the top level doesn't work, because it clashes with the module name, so remove the line from the __init__.py
pavelkomarov
committed

Commits on Jul 9, 2025

Commits on Jul 14, 2025

Commits on Jul 15, 2025