Nonlinear optimisation (root-finding, least squares, ...) in JAX+Equinox. https://docs.kidger.site/optimistix/
optimistix.{AbstractGaussNewton, GaussNewton, LevenbergMarquardt, IndirectLevenbergMarquardt, Dogleg}
should now all support using reverse-mode autodiff to calculate Jacobians. (#51)jax.disable_jit
(See https://github.com/patrick-kidger/diffrax/issues/368, #43)jax_numpy_rank_promotion=raise
and jax_numpy_dtype_promotion=strict
.Full Changelog: https://github.com/patrick-kidger/optimistix/compare/v0.0.6...v0.0.7
Bugfix release!
sol.state
previously included only the array-valued parts of the state (for sol = optx.{minimise, least_squares, root_find, fixed_point}(...)
). It now includes everything.optx.internal.implicit_jvp
misbehaving for non-Optimistix use cases.optx.{Newton,Chord}(cauchy_termination=False)
failing when started close to the solution.Full Changelog: https://github.com/patrick-kidger/optimistix/compare/v0.0.5...v0.0.6
Very simple release!
optimistix.compat.minimize
as a replacement for jax.scipy.optimize.minimize
. (#14)Full Changelog: https://github.com/patrick-kidger/optimistix/compare/v0.0.4...v0.0.5
Hurrah! How exciting.