v0.2.0

What's Changed

• Rename dense_output.py to solution.py by @pnkraemer in #480
• Group tests for IVP solvers and solutions by @pnkraemer in #481
• Move cubature module to implementations by @pnkraemer in #482
• Rename solvers by @pnkraemer in #483
• Renamed negative_marginal_log_likelihood to log_marginal_likelihood by @pnkraemer in #484
• Changelog by @pnkraemer in #485
• Replace "condition_on_qoi_observations" with "observe_qoi" by @pnkraemer in #487
• Conditionals by @pnkraemer in #488
• norm_of_whitened_residual_sqrtm is mahalanobis_norm by @pnkraemer in #491
• Moved test-dependencies to backend by @pnkraemer in #492
• Whitened residuals in IsoTS0 by @pnkraemer in #494
• Replaced jnp.sqrt(jnp.dot(...)) and jnp.sqrt(jnp.einsum(...)) with square-root arithmetic by @pnkraemer in #496
• Moved square-root utilities to top-level and renamed to sqrt_util by @pnkraemer in #497
• Sum of squares by @pnkraemer in #498
• Scalar solvers in public API by @pnkraemer in #499
• Doc improvements by @pnkraemer in #501
• Cubature factories by @pnkraemer in #502
• Skip JVP tests for Dense SLR1 by @pnkraemer in #503
• Removed unused method by @pnkraemer in #505
• IBM factories by @pnkraemer in #504
• Define public API by @pnkraemer in #506
• Correction factories by @pnkraemer in #507
• Remove type-checking tab by @pnkraemer in #510
• Implementation recipes are functions by @pnkraemer in #508
• Simplify extrapolation by @pnkraemer in #511
• Cache in state-space variable by @pnkraemer in #512
• Initial step in solver signature by @pnkraemer in #513
• Extract inverse to init for adaptive solvers by @pnkraemer in #514
• Time-step in controller state by @pnkraemer in #515
• Type-stability in filtering methods by @pnkraemer in #516
• The big output-scale-sqrtm overhaul by @pnkraemer in #517
• Initialise solution from posterior by @pnkraemer in #518
• Separate solver state from solution by @pnkraemer in #519
• Rename _ivpsolve_impl.py to _collocate.py by @pnkraemer in #520
• Group examples and benchmarks in directories by @pnkraemer in #521
• Remove 'rv1' from case_interpolate signature by @pnkraemer in #522
• Rename linearisation pt by @pnkraemer in #523
• Clean up strategy signatures by @pnkraemer in #524
• Use interpolation result type in strategies by @pnkraemer in #525
• Extract fn in strategy by @pnkraemer in #526
• Mention MLESolver fix in changelog by @pnkraemer in #527
• Improve usage of posterior vocabulary in solvers by @pnkraemer in #528
• Smoother state by @pnkraemer in #529
• Moved some properties from IVP solution to Strategy-states by @pnkraemer in #530
• Some cosmetic updates by @pnkraemer in #531
• Remove gain from complete_correction output by @pnkraemer in #532
• probdiffeq.implementations -> probdiffeq.statespace by @pnkraemer in #533
• Remove the middle-man: No more Implementation class. by @pnkraemer in #534
• Removed eq for solvers by @pnkraemer in #535
• init_output_scale is promote_output_scale by @pnkraemer in #536
• Move t field to strategy states by @pnkraemer in #537
• case_* methods private wherever appropriate by @pnkraemer in #538
• Removed _fn suffix for a bunch of methods by @pnkraemer in #539
• Extrapolation and correction define only begin and complete by @pnkraemer in #540
• Reintroduce mypy checks by @pnkraemer in #541
• Delete readthedocs file and format yaml files by @pnkraemer in #542
• Remove u and marginals from solver state by @pnkraemer in #543
• Moved sample() to solution by @pnkraemer in #544
• Mini updates by @pnkraemer in #545
• Remove unneded test cases by @pnkraemer in #547
• Increase fixed-grid test coverage by @pnkraemer in #548
• Increase save-at test coverage by @pnkraemer in #550
• Split statespace._collections.py file and abbreviate StateSpaceVar to SSV by @pnkraemer in #551
• Change function order in _collocate.py by @pnkraemer in #552
• Remove 'Abstract' from class names by @pnkraemer in #553
• Target shape in state-space variable by @pnkraemer in #554
• Increase terminal value test coverage by @pnkraemer in #549
• Filters and smoothers implement begin() and complete() directly by @pnkraemer in #555
• Filters implement only begin() and complete() now by @pnkraemer in #556
• Smoothers only implement begin and complete now by @pnkraemer in #557
• Strategy & statespace signatures by @pnkraemer in #558
• Remove num data points by @pnkraemer in #559
• Solution assembly outside of ivpsolvers.py by @pnkraemer in #560
• Extract output scale by @pnkraemer in #561
• Delete unused function by @pnkraemer in #562
• init_conditional -> smoother_init_conditional by @pnkraemer in #563
• Controls are now longer abstract by @pnkraemer in #564
• IVP solvers no longer abstract by @pnkraemer in #565
• Strategy no longer abstract by @pnkraemer in #566
• Removed unused caches in filter_begin and filter_complete by @pnkraemer in #567
• Remove _SmootherCommon by @pnkraemer in #568
• Fixed-point implementations gain dedicated begin/complete methods by @pnkraemer in #569
• Fix interpolation error by @pnkraemer in #571
• Test mle calibration by @pnkraemer in #570
• Moved u from _FiState and _SmState to SSV by @pnkraemer in #573
• Split extrapolations into filter-, smoother-, and fixedpoint-implementations by @pnkraemer in #574
• AdaptiveState is a named tuple now. by @pnkraemer in #575
• Extracted filtering functionality from LML by @pnkraemer in #576
• Dataclasses and Pytrees by @pnkraemer in #577
• Nested dataclasses by @pnkraemer in #579
• Removed custom dataclasses (they don't work as expected) by @pnkraemer in #580
• Phrased runge-kutta starter as a smoother by @pnkraemer in #581
• Removed last abstract classes by @pnkraemer in #582
• Split _collections into _markov and _interp by @pnkraemer in #583
• _collections, _conds, and _vars are all _vars now (in statespace/) by @pnkraemer in #584
• Remove init_error_estimate() by @pnkraemer in #585
• _vars -> variables by @pnkraemer in #586
• Removed SmootherSol and FilterDist by @pnkraemer in #587
• sqrtm_to_upper_triangular -> triu_via_qr by @pnkraemer in #588
• Use containers.NamedTuple instead of typing.NamedTuple by @pnkraemer in #589
• Overhauled the look of the docs by @pnkraemer in #590
• Predict, update, apply_updates scheme in solution._kalman_filter by @pnkraemer in #591
• Extract standard normal functionality by @pnkraemer in #592
• Moved up a variable in FP interpolation by @pnkraemer in #595
• Updated controller source-code by @pnkraemer in #594
• Renamed _duplicate_with_unit_backward_model to _reset_fixedpoint by @pnkraemer in #596
• Reset backward model in Fp.init by @pnkraemer in #593
• Merge backward-models in fixedpoint.complete() by @pnkraemer in #597
• Remove backward-model merging in case right corner by @pnkraemer in #598
• Simplify initialisation of proposed error estimate by @pnkraemer in #599
• Move merge_conditionals() to a function by @pnkraemer in #600
• Simplified extrapolation interface by @pnkraemer in #601
• Remove unused fields from AdaptiveState by @pnkraemer in #602
• Moved control flow to backend by @pnkraemer in #603
• Replace fixedpoint.init_conditional() with fixedpoint.reset() by @pnkraemer in #604
• Extracted init_conditional from dense extrapolation by @pnkraemer in #605
• Removed self.init_conditional() from Iso variables by @pnkraemer in #606
• Extracted identity conditional from scalar SSMs by @pnkraemer in #607
• Preconditioner is a Callable by @pnkraemer in #608
• Extract calibration from extrapolation by @pnkraemer in #609
• Move EOF and YAMLFIX to pre-commit hook. by @pnkraemer in #610
• Strategy as function by @pnkraemer in #611
• Calibration is now a property of the solver, not the strategy by @pnkraemer in #612
• IVP solvers are factory functions by @pnkraemer in #614
• Calibration methods are bundles of init-update-extract schemes by @pnkraemer in #615
• Call Calib.init() and Calib.extract() in IVP solver by @pnkraemer in #616
• BVP solver proof of concept by @pnkraemer in #617
• Replace mknotebooks with mkdocs-jupyter by @pnkraemer in #618
• Unify Markov sequence namings by @pnkraemer in #619
• Give test-suite a new look by @pnkraemer in #620
• Update action-setup versions by @pnkraemer in #624
• Increase python version to 3.9+ by @pnkraemer in #625
• Custom color scheme for Mkdocs by @pnkraemer in #626
• Update to changelogs by @pnkraemer in #627
• Remove scan_with_init and clean up Solution.getitem by @pnkraemer in #628
• Extract common post-processing functionality in ivpsolve by @pnkraemer in #629
• Implement ivpsolve.simulate_terminal_values via _collocate.solve_and_save_at by @pnkraemer in #630
• Move scale_covariance logic to ivpsolve.py by @pnkraemer in #631
• Move interpolation-logic to collocate.py by @pnkraemer in #632
• Simplify interpolation interpolation via moving logic to _collocate.py by @pnkraemer in #633
• Unify the implementations in strategies by @pnkraemer in #634
• Compositional implementation of strategies by @pnkraemer in #635
• Compositional implementation of ivpsolvers by @pnkraemer in #636
• Factory methods in extrapolation-model construction by @pnkraemer in #637
• Calibration factory by @pnkraemer in #638
• Move solution object from solution.py to ivpsolve.py by @pnkraemer in #639
• Separate ivpsolver package by @pnkraemer in #640
• Make strategies.strategy public by @pnkraemer in #641
• Clean up solver API, calibration implementation; make output_scale compulsory by @pnkraemer in #642
• Controller constructors by @pnkraemer in #643
• Tightened the API for Correction models by @pnkraemer in #644
• Correction model updates by @pnkraemer in #645
• Iso-factorisation implements the new correction-model style by @pnkraemer in #646
• Block-diag implementations use new API by @pnkraemer in #647
• Scalar solvers implement corrections by @pnkraemer in #648
• Clean up extrapolation factory by @pnkraemer in #649
• Compulsory output scale in ibm_utils.system_matrices by @pnkraemer in #650
• Removed unused arguments in right-corner function by @pnkraemer in #651
• Towards v0.2.0 by @pnkraemer in #654

Full Changelog: v0.1.4...v0.2.0

See https://pnkraemer.github.io/probdiffeq/getting_started/easy_example.md for an introduction to the new API .