1.2.4

What's Changed

• Simplify run_inference_algorithm by @reubenharry in #714
• Bug: Harmonize doc and README code examples by @gil2rok in #717
• Fix README example by @gil2rok in #719
• Bug fix by @reubenharry in #724
• Waste Free SMC available for adaptive tempered and tempered SMC. by @ciguaran in #721
• NaN Handling by @reubenharry in #727
• Implement metric scaling by @AdrienCorenflos in #733
• Speed up Schrodinger Follmer test by @AdrienCorenflos in #741

Full Changelog: 1.2.3...1.2.4

1.2.3

What's Changed

• Fixing typo in example on index page - leading to 30x faster runtime. by @johannahaffner in #711
• Enable progress bar under pmap by @andrewdipper in #712
• wrapper for progress _bar_scan by @andrewdipper in #716

New Contributors

• @johannahaffner made their first contribution in #711

Full Changelog: 1.2.2...1.2.3

1.2.2

What's Changed

• Make tests pass on aarch64-linux by @albcab in #671
• Enable fitlering of AdaptationInfo by @andrewdipper in #674
• Update run_inference_algorithm to split initial_position and initial_state by @reubenharry in #672
• Preconditioned mclmc by @reubenharry in #673
• New integrator, and add some metadata to integrators.py by @reubenharry in #681
• Minor formatting by @junpenglao in #685
• Make Window Adaptation Take Integrator As Argument by @reubenharry in #687
• FIX KWARG BUG by @reubenharry in #686
• Change isokinetic_integrator generation API by @junpenglao in #689
• Apply function on pytree directly. by @junpenglao in #692
• Fix sampling test. by @junpenglao in #693
• Enable shared mcmc parameters with tempered smc by @andrewdipper in #694
• Convert explicit looping to bit twiddling for iterative_uturn calculations by @andrewdipper in #696
• Remove nightly release by @junpenglao in #699
• Fix doc mistakes by @gil2rok in #701

New Contributors

• @andrewdipper made their first contribution in #674
• @gil2rok made their first contribution in #701

Full Changelog: 1.2.1...1.2.2

1.2.1

What's Changed

• Fix deprecated call to jnp.clip by @GaetanLepage in #664
• Update jax version requirements by @junpenglao in #666

New Contributors

• @GaetanLepage made their first contribution in #664

Full Changelog: 1.2.0...1.2.1

Blackjax v1.2.0

Mostly internal change, a major one being a different implementation for top level api (#657)

What's Changed

• Indexing the notebook showing how to reproduce the GIF. by @AdrienCorenflos in #640
• Bump python version by @junpenglao in #645
• SMC: allow each mutation kernel to have different parameters. by @ciguaran in #649
• Migrate from deprecated host_callback to io_callback by @junpenglao in #651
• Fix MALA transition energy by @ksnxr in #653
• Change variable names by @ksnxr in #654
• Replace iterative RNG split and carry with jax.random.fold_in by @junpenglao in #656
• Removal of Algorithm classes. by @ciguaran in #657

New Contributors

• @ksnxr made their first contribution in #653

Full Changelog: 1.1.1...1.2.0

1.1.1

What's Changed

• minor clean up by @junpenglao in #635
• Example showing how to use BlackJAX to sample from the word "BlackJAX" by @AdrienCorenflos in #637
• Update README.md by @AdrienCorenflos in #638

Full Changelog: 1.1.0...1.1.1

Blackjax v1.1.0

New Features Highlights

• ChEES-HMC
• MCLMC sampler
• Riemannian Manifold HMC sampler

What's Changed

• Add Governance and Code of Conduct by @junpenglao in #570
• bump github action version by @junpenglao in #573
• Documentation clean up by @junpenglao in #572
• Check only for negative divergences by @jdehning in #574
• Update requirements-doc.txt by @junpenglao in #578
• Refactor divergence check to each sampler by @junpenglao in #579
• Dynamic HMC by @albcab in #580
• Adjusts for the fact that IRMH proposal might not be symmetric by @ciguaran in #581
• Document how particles should be represented in SMC by @ciguaran in #582
• Add adaptation on HMC trajectory length to increase the expected Change in the Estimator of the Expected Square (ChEES-HMC) by @albcab in #421
• Update chees_adaptation.py by @junpenglao in #583
• ghmc dev by @williwilliams3 in #585
• EuclideanIntegrator -> Integrator by @albcab in #588
• Refactor integrators.py to make it more general. by @junpenglao in #589
• Allow passing line search algorithm to pathfinder. by @ColCarroll in #592
• Name arguments consistently across adaptation classes. by @ColCarroll in #593
• Improve typing and doc of proposal generation by @junpenglao in #594
• Implements Barker's proposal. by @AdrienCorenflos in #596
• Add Barker's proposal reference by @AdrienCorenflos in #600
• Implements the Schrödinger-Föllmer algorithm. by @AdrienCorenflos in #602
• Refactor proposal.py by @junpenglao in #603
• Add MCLMC sampler by @reubenharry in #586
• Losen the input requirement for ess computation. by @junpenglao in #609
• SMC Inner kernel tuning by @ciguaran in #595
• Resolve 'Functions to run kernels' by @PaulScemama in #598
• Minor clean up for SMC by @junpenglao in #611
• MInor docstring fix by @junpenglao in #612
• Docstring fix by @junpenglao in #613
• Resolve: Add progress bar to run_inference_algorithm by @PaulScemama in #614
• Add transform TO run_inference_loop by @reubenharry in #621
• Remove transform from MCLMC by @reubenharry in #623
• Refactor dynamic_hmc out of hmc.py by @junpenglao in #622
• Clean up of run_inference_algorithm docstring. by @junpenglao in #624
• Update noneuclidean integrator by @junpenglao in #626
• Refactor mgrad_gaussian by @junpenglao in #628
• Extend function signature for InitFn by @junpenglao in #627
• Adding Riemannian Manifold HMC by @dfm in #538
• RENAME NONEUCLIDEAN TO ISOKINETIC by @reubenharry in #632
• PREVENT 1D DISTRIBUTION by @reubenharry in #633

New Contributors

• @jdehning made their first contribution in #574
• @williwilliams3 made their first contribution in #585
• @ColCarroll made their first contribution in #592
• @reubenharry made their first contribution in #586
• @PaulScemama made their first contribution in #598

Full Changelog: 1.0.0...1.1.0

Blackjax v1.0.0

What's Changed

• Refactor the adaptation kernels by @rlouf in #276
• Update change_of_variable_hmc.md by @junpenglao in #373
• Remove logprob_grad_fn kwargs in favor of jax.custom_vjp by @rlouf in #342
• Fix benchmark by @junpenglao in #380
• Refactor the function for generating Gaussian noise by @junpenglao in #377
• Improve batch shape handling in MEADS and allow scaled step sizes in GHMC using mcmc.integrators.velocity_verlet by @albcab in #381
• Modify linear regressions to parameters in R by @albcab in #388
• S/acceptance probability/acceptance ratio by @rlouf in #390
• Update the pathfinder API by @rlouf in #392
• Add the control variates gradient estimator by @rlouf in #299
• Make tests fail on uncaught warnings by @rlouf in #404
• Update the issue templates by @rlouf in #409
• Add the licence notice at the top of every file by @rlouf in #408
• Clean examples by @rlouf in #413
• Add linebreak between function params in API doc by @rlouf in #414
• Add support for nightly releases by @rlouf in #418
• Use NumPy style for linebreak at end of docstring by @rlouf in #419
• Use the new marginal sampler for latent gaussian models in the GP notebook by @juanitorduz in #398
• Fix import for blackjax.optimizers.dual_averaging by @junpenglao in #426
• Fix pymc example by @zaxtax in #424
• Minor doc clean up by @junpenglao in #428
• s/logprob/logdensity by @junpenglao in #427
• Use pyproject.toml and setuptools_scm by @rlouf in #431
• Gitlint configuration by @valentynbez in #429
• Add Contour SGMCMC sampler. by @WayneDW in #396
• Automatically open an issue for Blackjax meetings by @rlouf in #440
• Add ReadTheDocs configuration file by @rlouf in #432
• Add Mean Field Variational Inference implementation by @xidulu in #433
• Fix small typos by @antotocar34 in #460
• Refactor the potential fun flip in HMC by @junpenglao in #463
• Move Contour SGLD example to the Sampling Book by @rlouf in #467
• Simplify the Oryx examples by @rlouf in #468
• Remove the hierarhical BNN example by @rlouf in #469
• Return adaptation extra information by @rlouf in #466
• Refactor the SMC kernels by @rlouf in #279
• Remove the sparse logistic regression example by @rlouf in #472
• Fixing TemperedSMC example by @ciguaran in #474
• Add documentation for how to implement Metropolis-within-Gibbs by @mlysy in #422
• Use github-action-benchmark for continous benchmarking by @rlouf in #476
• Fix the Introduction Example by @yayami3 in #477
• Remove unnecessary dependency by @yayami3 in #481
• Replace uses of internal JAX NumPy utils with public API functions. by @junpenglao in #487
• Adding folder structure to tests by @ciguaran in #483
• Remove the remaining examples by @rlouf in #489
• Update pre-commit dependencies by @rlouf in #490
• Use AutoAPI for the API documentation by @rlouf in #478
• Handle references using sphinxcontrib-bibtex with the BibTeX alpha style by @albcab in #494
• Use public JAX API for PRNGKeyArray by @junpenglao in #498
• Do not fail RTD build on warning by @rlouf in #502
• use chex for typing by @junpenglao in #503
• Exposing RMH and Random Walk as two different algorithms, generalizing RW to non-gaussian jumps by @ciguaran in #484
• Drop support of Python 3.7 and install latest developments of optax to avoid failing tests by @albcab in #519
• Pass grad_estimator to the CSGLD kernel directly by @albcab in #518
• Fix SGLD and SGHMC docstrings by @albcab in #524
• Refactor MALA so that it uses the MH component in proposals.py by @albcab in #523
• SMC-MCMC integration test, plus fixes. by @ciguaran in #522
• Use latest release of optax by @albcab in #528
• Revert "Change documentation background color" by @howsiyu in #527
• Proper splitting of PRNG key by @howsiyu in #526
• Consistent naming of low-level function by @albcab in #532
• Fix momentum samples for 2D inverse mass matrix by @junpenglao in #533
• Move kernel functions to their algorithm-class folder. by @albcab in #501
• Add SGNHT by @SamDuffield in #515
• Converting scan to fori_loop in static_integration in order to support a variable number of integration steps in HMC sampler by @dfm in #539
• Implement SVGD by @antotocar34 in #512
• Add utility function to demonstrate unflattening of arrays by @albcab in #535
• Unify SGMCMC init structure by @SamDuffield in #540
• Change Typing to follow Jax best practice by @junpenglao in #543
• MCMCSamplingAlgorithm -> SamplingAlgorithm by @albcab in #542
• Update howto_use_tfp.md by @junpenglao in #553
• Documentation clean up by @junpenglao in #552
• Bump to Py3.9 by @junpenglao in #554
• Update usage of soon-to-be deprecated Chex assertions by @junpenglao in #560
• Moving max_num_doublings of NUTS from build_kernel to kernel by @weiyaw in #566
• Update the Metropolis-within-Gibbs example by @twhentschel in #558
• Update dependencies and use new jax random key by @junpenglao in #569

New Contributors

• @juanitorduz made their first contribution in #398
• @valentynbez made their first contribution in #429
• @WayneDW made their first contribution in #396
• @xidulu made their first contribution in #433
• @antotocar34 made their first contribution in #460
• @mlysy made their first contribution in #422
• @yayami3 made their first contribution in #477
• @howsiyu made their first contribution in #527
• @SamDuffield made their first contribution in #515
• @dfm made their first contribution in #539
• @weiyaw made their first contribution in #566
• @twhentschel made their first contribution in #558

Full Changelog: 0.9.6...1.0.0

Blackjax v0.9.6

What's Changed

• Add How-to reference on using logpdf built with another framework by @rlouf in #345
• Fix documentation typo for kernel and state input by @oarriaga in #367

New Contributors

• @oarriaga made their first contribution in #367

Full Changelog: 0.9.5...0.9.6

Blackjax v0.9.5

What's Changed

• Refactor MNIST example to use Flax by @zaxtax in #365

Full Changelog: 0.9.4...0.9.5