Skip to content

Blackjax v1.0.0
Compare
Choose a tag to compare
@junpenglao junpenglao released this 20 Sep 19:35
· 87 commits to main since this release
1.0.0
8efb158
This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Learn about vigilant mode.

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

Full Changelog: 0.9.6...1.0.0

Contributors

zaxtax, dfm, and 15 other contributors
Assets 2
Loading