Per-parameter scaling via change of variables z = x / D. Supports
'jac' (adaptive D_i = 1/||J(:,i)|| per iteration, matches scipy's TRF),
explicit array, or a positive scalar. Default 1.0 is a no-op.
All changes make functionality more similar to SciPy least squares:
- Use adaptive findiff epsilon.
- Make termination on step size relative to norm(x).
- Add termination on gradient norm.
- Make default tolerances like SciPy's.
PiperOrigin-RevId: 745221614
Change-Id: Iee93256651fca8154c97fa3bdaa9c67ede28e573
- Residual callable is vectorized for easy multithreading by the user. Internally all vectors are now explicitly column vectors.
- Removed central findiff option, it wasn't applicable at the bounds anyway and just complicated the code.
- Added optional user-provided norm function for non-quadratic (robust) norms.
- Added an iter_callback callable for user convenience.
- Added option to internally check user-provided Jacobian and norm against finite differences.
- Updated the notebook accordingly.
PiperOrigin-RevId: 631205889
Change-Id: I3b9f8893756e329640de464e6f2e26a39e7dbd9e
- Added central differencing option.
- Changed `mu` semantics from log10 to actual values.
- `mu` control is now more aggressive, saves a few iterations.
- Use Armijo sufficient reduction criterion.
- Added some checks for float64 and NaNs.
- Added log of final result to trace.
- Removed unhelpful QP timing.
PiperOrigin-RevId: 610096950
Change-Id: Idf1a7c3155ce8e82fc2d5fe0f2da2f16636ff63d
Copybara-Service