Release v1.5.3

Bugfix when starting solver at problem minimizer

Release v1.5.2

• Bugfix for saving diagnostic info (bug introduced in v1.5.1), issue #23

Release v1.5.1

• Add new return values soln.xmin_eval_num and soln.jacmin_eval_nums
• Allow option for parallel initial evaluations for non-random directions if npt not too large

Neither should affect the performance of the DFO-LS algorithm

Release v1.5.0

Allow optional (nonsmooth) regularization term in the objective to avoid overfitting.

Also DFO-LS now does not give warnings if the initial point is on the upper/lower bounds (only if it is outside the bounds).

Release v1.4.1

Migrate installation to pyproject.toml (required for Python 3.12+) and drop support for Python 2.7 and <=3.8 to align with SciPy >=1.11 dependency. No change to the expected behavior of the DFO-LS algorithm.

Release v1.4

Upgraded required scipy version and graceful handling of NaNs in objective evaluation in trust-region step. No change to expected behavior of DFO-LS algorithm.

Release v1.3

Major update: DFO-LS can now handle arbitrary convex constraints, supplied as a list of projection operators. No change to unconstrained and bound constrained solver.

Release v1.2.3

Minor update to customise handling of NaNs in objective evaluations - no changes to the DFO-LS algorithm.

Release v1.2.2

Minor update to remove NumPy deprecation warnings - no changes to the DFO-LS algorithm.

Release v1.2.1

Minor release: dependency on trustregion package now optional. DFO-LS can use the trustregion package if it is installed, but will still work if the it is not installed (and DFO-LS will not install it by default).