TuringLang/Turing.jl: v0.22.0

Turing v0.22.0

Diff since v0.21.13

Closed issues:

• [BNP] Bayesian Nonparametrics project (#370)
• [BNP] Q-class (#374)
• [MH] change of variable (#376)
• [Traces]: add a member field to Trace to mark reference particle (#396)
• More robust mechanism for mapping model variables to samplers (#397)
• Particle Gibbs does not return log probability (#643)
• Improvement of sampler descriptions. (wip) (#699)
• Test for RandomMeasures fail (#973)
• Update benchmarks on wiki (#1006)
• Sanity check is missing for space of Gibbs sampler and model parameter space (#1012)
• Handle gradient = nothing case when doing reverse diff with Zygote (#1081)
• Failure when dimension changes (#1115)
• Simple benchmarking for Tracker, ReverseDiff, and Zygote (#1140)
• initial parameters / NUTS step size (#1314)
• Integrating out variables via Laplace approximation (or other methods) (#1382)
• Add instructions to check the gradient function (#1383)
• Precompilation fails while using ReverseDiff backend (#1400)
• RFC: Sampler-specific AD settings instead of global AD settings (#1402)
• rdcache not cleared while using MLE,MAP estimation (#1418)
• Coeftable throws error on mle and map estimates (#1447)
• GibbsConditionals: proper variable subsumption (#1452)
• Wrong interpolation symbol? (#1551)
• Sampler initialization (init_params) (#1563)
• Some tests for MH fail for certain random seeds (#1587)
• The chains do not converge to the same posterior (#1593)
• Numerical error messages: real problem or just terminal noise? (#1621)
• Feature request: Add LKJCholesky (#1629)
• Sometimes this error is thrown: Cannot convert an object of type Geometric{Float64} to an object of type Dirac{Float64} (#1669)
• Improper (?) chains resuming (#1697)
• AbstractMCMC.step with different model at each step (#1699)
• Stein Thinning (#1704)
• Question: looks like the order of variables are inconsistent (#1712)
• Bijector error with VI and arraydist/filldist (#1717)
• Easy way to get gradient evaluation timing (#1721)
• Automatic lazy broadcasting/optimization for arrays of distributions (#1723)
• Invalid test case for Gibbs? (#1725)
• Refactoring Glue Code (#1735)
• Way to Declare Data (#1736)
• Improve usage and support for immutable AbstractVarInfo (#1752)
• Batch size reduced to zero when doing inference on ODEs (#1760)
• Massive test time regressions (#1774)
• Broken doctests (#1796)
• there is an erro in Variational inference (VI) of the Tutorials (#1820)
• Concrete terse structs (#1830)
• Question: Which package is the MCMCserial() part of? been working through the notebook and have had problems with it not being defined hence that part of the notebook does not run (#1842)
• Custom distribution sampling does not perform well (#1859)
• Proposal: Interest in faster logpdf for BernoulliLogit (logistic regression) (#1890)

Merged pull requests:

• Compatibility with new DPPL version (#1900) (@torfjelde)
• CompatHelper: bump compat for Setfield to 1, (keep existing compat) (#1906) (@github-actions[bot])
• Bump minor version (#1909) (@torfjelde)