v2.0.1

EnsembleKalmanProcesses v2.0.1

Diff since v2.0.0

Merged pull requests:

• remove warning for now on Accelerator vs deterministic_forward_map (#400) (@odunbar)
• Update Project.toml (#401) (@odunbar)

Closed issues:

• Conflicting default with warning message (#399)

v2.0.0

EnsembleKalmanProcesses v2.0.0

Diff since v1.1.7

The primary breaking changes of this are:

• We have a new philosphy of defaults. Before this meant the featureless "vanilla" process; Now this means including the "latest" stable features. The default configurations are detailed in the docs here
• The new lego-like observations and minibatching framework, detailed in the docs here
• Many patches to the interface to implement and export getter methods get_XYZ(ekp) replacing ekp.XYZ for internals

Merged pull requests:

• More flexible Interpolations dependency (#391) (@odunbar)
• Replace vanilla defaults with current "cutting-edge" settings (#392) (@odunbar)
• Docs/Readme for defaults and recommendations (#394) (@odunbar)
• Miscellaneous interface changes, solving several older issues (#395) (@odunbar)
• updated accelerator examples (#397) (@odunbar)
• Update Project.toml to v2.0.0 breaking (#398) (@odunbar)

Closed issues:

• obs_mean has no type, yet later is assumed a vector (#263)
• Add ETKI into the landing page. (#352)
• transform_unconstrained_to_constrained behavior is unclear for single values and Vectors (#369)
• Link docs to example folder (#371)
• O3.2.8 EKP updates from the Migration TC-EDMF calibration pipeline into CalibrateAtmos/ClimaAtmos (#373)
• get_\phi should have the return_array keyword (#381)
• O3.2.4 Hardcode sensible defaults (#387)
• Accelerator examples must set default (#396)

v1.1.7

EnsembleKalmanProcesses v1.1.7

Diff since v1.1.6

Merged pull requests:

• Bump version to v1.1.7 after bugfix (#390) (@odunbar)

Closed issues:

• Wrong type for observation (#389)

v1.1.6

EnsembleKalmanProcesses v1.1.6

Diff since v1.1.5

Merged pull requests:

• Fix a typo in darcy.md (#346) (@glwagner)
• remove positive definiteness constraints, allow user defined additive inflation (#360) (@odunbar)
• CompatHelper: bump compat for SCS to 2, (keep existing compat) (#361) (@github-actions[bot])
• add Project.toml for Localization example (#362) (@odunbar)
• bugfix logpdf broadcasting (#364) (@odunbar)
• NICE sample-error correction (#367) (@odunbar)
• Add troubleshooting doc (#368) (@costachris)
• Add save_parameter_samples (#370) (@nefrathenrici)
• CompatHelper: add new compat entry for Interpolations at version 0.15, (keep existing compat) (#376) (@github-actions[bot])
• CompatHelper: bump compat for Convex to 0.16, (keep existing compat) (#379) (@github-actions[bot])
• Complete redesign of "Observations" object enabling introduction of minibatching (#384) (@odunbar)
• Update version to v1.1.6 (#388) (@odunbar)

Closed issues:

• O3.7.3 Overcome precompiling every (julia) ensemble member on HPC (#331)
• No Project.toml for the Localization example (#358)
• Positive definite corrections in get_u_cov (#359)
• Remove broadcasting for Logpdf. (#363)
• O3.7.7 Design a user-friendly guide for configuring EnsembleKalmanProcess (#365)
• Make SECFisher more accurate (#366)
• Improve speed of SECNice (#372)
• Add convenient method for minibatching data (#382)
• Add ability to mutate key quantities such as the observation covariance matrix (#383)
• ETKI ignores timestepper (#385)

v1.1.5

EnsembleKalmanProcesses v1.1.5

Diff since v1.1.4

Merged pull requests:

• Add examples for accelerators (#336) (@sydneyvernon)
• Updated accelerators with variable timestep, add localization case to examples (#348) (@odunbar)
• Handling bors removal: new workflows (#349) (@odunbar)
• Remove bors.toml (#350) (@odunbar)
• Update README.md (#351) (@odunbar)
• Add constrained_gaussian TOML parser (#353) (@nefrathenrici)
• Remove getfield(Main, ...) (#355) (@nefrathenrici)
• Update Project.toml to v1.1.5 (#357) (@odunbar)

Closed issues:

• Add parser for constrained_gaussian from TOML (#314)
• Implement Example for some nonlinear problems (#340)
• getfield of Main (#354)

v1.1.4

EnsembleKalmanProcesses v1.1.4

Diff since v1.1.3

Merged pull requests:

• Documentation for ETKI (#337) (@eviatarbach)
• Prevent DataMisfitController scheduler from modifying EKP obs_noise_cov (#338) (@costachris)
• modified accelerator for variable timestep (#341) (@odunbar)
• CompatHelper: add new compat entry for TOML at version 1, (keep existing compat) (#342) (@github-actions[bot])
• CompatHelper: add new compat entry for Statistics at version 1, (keep existing compat) (#343) (@github-actions[bot])
• Update project to v1.1.4 (#344) (@odunbar)
• Update to new [compat] requirements (#345) (@odunbar)

Closed issues:

• O3.7.2 Implement ETKI for high-dimensional output scaling benefits (#299)
• Examples and documentation for ETKI (#330)
• Modfiy to work with variable timestepping (#339)

v1.1.3

EnsembleKalmanProcesses v1.1.3

Diff since v1.1.2

Merged pull requests:

• Darcy example documentation (#318) (@odunbar)
• WIP: adding momentum-inspired accelerators to EKP (#322) (@sydneyvernon)
• Bugfix darcy plot (#326) (@odunbar)
• docs update (#327) (@odunbar)
• docs update (#328) (@odunbar)
• Ensemble tranform Kalman inversion (#329) (@eviatarbach)
• Unscented accelerator (#332) (@odunbar)

Closed issues:

• Document Darcy example (#317)
• Implement Source code and Unit tests in a method-independent way (#323)
• Implement and test ensemble transform Kalman inversion (ETKI) (#324)
• darcy.jl: problem when plotting true flow (#325)
• Implementation of UKI acceleration based on sigma point movement (#333)
• Soft-sparsity localization of C^ug (#334)

v1.1.2

EnsembleKalmanProcesses v1.1.2

Diff since v1.1.1

Closed issues:

• Parameter Distributions is not well compatible with matrix distributions (#229)
• remove cat(X...) statements (#274)
• UKI improperly tested - and other test issues... (#287)
• DataMisfitController() == DataMisfitController() gives false (#297)
• Base.show(::ParameterDistribution) error with Gaussian Random Fields (#300)
• Compatibility issues while building deployment docs (#305)
• Tighten terminology in docs (#308)
• Learning rate schedulers code snippet in documentation (#311)
• Covariance Localization using Ledoit Wolf (#313)

Merged pull requests:

• Fast concatentation (#283) (@haakon-e)
• Resolving some test problem consistency and apparent UKI instability (#294) (@odunbar)
• Add == for schedulers (#298) (@odunbar)
• minor typo fix for docstring of get_ϕ (#302) (@wilsonduan10)
• Function parameter documentation (#304) (@odunbar)
• Fixing small error in example documentation (#309) (@eviatarbach)
• Update learning_rate_scheduler.md (#310) (@eviatarbach)
• Removing unused history field in DataMisfitController (#312) (@eviatarbach)
• Small docs updates (#315) (@odunbar)

v1.1.1

EnsembleKalmanProcesses v1.1.1

Diff since v1.1.0

Closed issues:

• Functionality to plot prior pdf (in constrained and unconstrained space) (#52)
• Converting non-flattened parameter arrays (#130)
• Small note to add to docs (#226)
• We can improve DMC consistency by controlling the termination tolerance (#290)
• Adaptive schedulers not compatible with particle failures (#292)

Merged pull requests:

• CompatHelper: bump compat for StatsBase to 0.34, (keep existing compat) (#284) (@github-actions[bot])
• Allow users to select termination time for DataMisfitController (#291) (@odunbar)
• Recommended Constructor Docs update (#293) (@odunbar)
• Schedulers no longer break when particle failures occur (#295) (@odunbar)

v1.1.0

EnsembleKalmanProcesses v1.1.0

Diff since v1.0.0

Closed issues:

• Add LearningRateSchedulers (#154)
• Gaussian Random Field parameters (#251)
• Timestepping in EKP (#253)
• Should be passing an RNG here, not just a seed (#260)
• EKP for general likelihoods? (#262)
• Remove use of covariance determinant (#271)
• License copyright text not filled out (#276)

Merged pull requests:

• Add GaussianRandomFields to ParameterDistributions (#252) (@odunbar)
• CompatHelper: add new compat entry for GaussianRandomFields at version 2, (keep existing compat) (#275) (@github-actions[bot])
• Update LICENSE (#277) (@odunbar)
• license (#278) (@odunbar)
• removing find_ekp_stepsize (#280) (@eviatarbach)
• pass RNG instead of seeds (#281) (@costachris)
• Timesteppers for EKP, from ParameterEstimocean (#285) (@odunbar)
• Fix v1.9.0 testing (#286) (@odunbar)