Recent Releases of Unfold
Unfold - v0.8.5
Unfold v0.8.5
- new function
StatsAPI.r2calculating coefficient of determination - new tutorial how to exclude bad data from model-fit
- 🐍 documentation fixes
Merged pull requests: - CompatHelper: bump compat for BSplineKit to 0.19, (keep existing compat) (#259) (@github-actions[bot]) - CompatHelper: bump compat for BSplineKit in [weakdeps] to 0.19, (keep existing compat) (#260) (@github-actions[bot]) - [AUTO] pre-commit update (#261) (@github-actions[bot]) - CompatHelper: bump compat for Krylov in [weakdeps] to 0.10, (keep existing compat) (#262) (@github-actions[bot]) - CompatHelper: bump compat for Krylov to 0.10, (keep existing compat) (#263) (@github-actions[bot]) - Unify landingpage - Update doc links to the respective UnfoldDocs pages. (#264) (@xuyg16) - new tutorial, addressing #270 (#271) (@behinger) - feature r2 (#272) (@behinger) - CompatHelper: bump compat for Interpolations to 0.16, (keep existing compat) (#273) (@github-actions[bot]) - [AUTO] pre-commit update (#274) (@github-actions[bot]) - docs: add LeonvonHaugwitz as a contributor for bug (#278) (@allcontributors[bot])
Closed issues: - differentiate between sparse and non sparse basisfunctions (#11) - MultiDocumenter.jl (#207)
- Julia
Published by github-actions[bot] 8 months ago
Unfold - v0.8.2
Unfold v0.8.2
- Update docstring of condense function
- Updated some tutorials + better looking docs
Merged pull requests: - Update condense.jl (#245) (@ReneSkukies) - [AUTO] pre-commit update (#246) (@github-actions[bot]) - CompatHelper: bump compat for BSplineKit to 0.18, (keep existing compat) (#248) (@github-actions[bot]) - CompatHelper: bump compat for BSplineKit in [weakdeps] to 0.18, (keep existing compat) (#249) (@github-actions[bot]) - removed Unfold from test environment... (#250) (@behinger) - [AUTO] pre-commit update (#251) (@github-actions[bot]) - fix: BSPlineKit 0.18 is not breaking, but now exports fit, which runs into a conflict (#252) (@behinger)
- Julia
Published by github-actions[bot] 11 months ago
Unfold - v0.8.0
Unfold v0.8.0
breaking: Instead of having MixedModels as a PackageExtension, we will soon have the UnfoldMixedModels.jl package (after 3 days of registry period). This is necessary to export the UnfoldLinearMixedModels type, which is not exprtable from a PackageExtension. This split further allows for better code modularity.
That is, users now have to do: using UnfoldMixedModels rather than using Unfold, MixedModels
Merged pull requests: - CompatHelper: bump compat for DSP to 0.8, (keep existing compat) (#237) (@github-actions[bot]) - [AUTO] pre-commit update (#238) (@github-actions[bot]) - predict docs typo fix (#240) (@Till223) - docs: add Till223 as a contributor for doc, and bug (#241) (@allcontributors[bot]) - remove @show from epoch.jl (#242) (@behinger) - (re)moved MixedModelsExt traces (#243) (@behinger)
Closed issues: - epoch() contains an @show (#239)
- Julia
Published by github-actions[bot] about 1 year ago
Unfold - v0.7.8
Unfold v0.7.8
Feature:
- Scaled FIR basisfunction ala Cameron Hassall et al. 2022 implemented
- Scaled FIR basisfunction ala Blockdesign
[0 0 0 1 1 1 1 1 0 0]implemented - New tutorial for those # Backend
- Moved all actions to
BestieTemplates.jl, including online Formatter action to pre-commit
Merged pull requests: - FIR with duration (#89) (@behinger) - apply BestieTemplate - :S (#229) (@behinger) - Bump codecov/codecov-action from 4 to 5 (#231) (@dependabot[bot]) - Bump actions/upload-artifact from 2 to 4 (#232) (@dependabot[bot]) - Bump actions/checkout from 2 to 4 (#233) (@dependabot[bot]) - Bump peter-evans/create-or-update-comment from 3 to 4 (#234) (@dependabot[bot]) - Bump peter-evans/find-comment from 2 to 3 (#235) (@dependabot[bot])
- Julia
Published by github-actions[bot] about 1 year ago
Unfold - v0.7.7
Unfold v0.7.7
- new structured backbone for the solver. Added ~6-7 new solvers. Split up the solver into prepare & solve functions that can be chained / modified easily. Added a reference document for exactly this.
- major bug: likelihoodratiotest was calculating wrong results (it was calculating a wrong likelihood). This lead to a breaking change in API, where now the data need to be provided as well.
- Bumped JLD to 0.5 - this means one cannot save files now and load it in older Unfold versions, but old unfold version files can still be loaded
- several small docstring fixes
- new CUDAextension, allowing to provide CuArray to automatically choose a better solver (still needs to be tested more though)
- fit now has an argument that can be set fit=false to not actually run the ´fit!` function, to do it later
- benchmarks now available comparing ~7 solver types with and without GPU, some with and without Float32. Improvements of up to 10x of the SOA-GPU solver were achieved - fit a large model with 128 channels in wopping 8 seconds! Further tests are needed though to describe the accuracy drop that is achieved by solving
X'Xb = X'yinstead ofXb = ydirectly.
Merged pull requests: - predict tutorial (#219) (@behinger) - better fit docs (#221) (@behinger) - docs: add maxvanmigem as a contributor for bug (#222) (@allcontributors[bot]) - update to JLD 0.5 (#224) (@behinger) - docs: add ReneSkukies as a contributor for code, and ideas (#226) (@allcontributors[bot]) - new solver system (#227) (@behinger) - better benchmark structure (#228) (@behinger)
Closed issues: - documentation open issues (#195) - LM overlap correction tutorial still uses old Unfold syntax (#199) - Inconsistency in predict( ) input; better docstring (#216) - Julia 1.11 crash (#223) - slight differences between MixedModels.jl and Unfold.jl (#225)
- Julia
Published by github-actions[bot] over 1 year ago
Unfold - v0.7.6_zenodo
new release specifically to fix a bug with zenodo
- Julia
Published by behinger over 1 year ago
Unfold - v0.7.6
Unfold v0.7.6
- bugfix for predict @ReneSkukies
- small documentation fixes
- fix Zenodo release metadata
Merged pull requests: - Update juliacallunfold.jl to include MNE Python warning (#211) (@carmenamme) - docs: add carmenamme as a contributor for doc (#212) (@allcontributors[bot]) - Remove spaces between docstring and effect function (#215) (@jschepers) - Update predict (#217) (@ReneSkukies) - Grammar and readability for nonlineareffects.jl (#218) (@vladdez)
- Julia
Published by github-actions[bot] over 1 year ago
Unfold - v0.7.3
Unfold v0.7.3
- fixed major bug in UnfoldLinearModel introduced in 0.7.0. If you used for MassUnivariate models with
Anythis bug did not exist. But if you subset your data e.g. via[:fixation=>(@formula(0~1),times)]or similar then you have to recalculate your results! In this case the subsetting of the data/epochs went wrong due to a nasty, but very tiny bug. ASubDataFramewas silently cast to aDataFrame, removingthe information which events belonged to which index.- Added small feature to be able to call
effects(Dict(),model)which adds a dummy effect to the dict, to be able to get the pure intercept.
- Added small feature to be able to call
Merged pull requests: - emptyeffects (#198) (@behinger) - fix mass univariate bug (#205) (@behinger)
Closed issues: - Allow empty dict in effects (#167)
- Julia
Published by github-actions[bot] over 1 year ago
Unfold - v0.7.2
Unfold v0.7.2
- bugfix for missing data & krylov solver
- bugfix with non-integer event onsets #200
- new internal functionality
_residualsnow allows to calculate residuals directly cut to an event, not requiring to generate the wholey = X*b
Merged pull requests: - fix200 (#201) (@behinger) - fixMissing (#202) (@behinger) - residual (#203) (@behinger)
Closed issues: - Bug: basisfunction with non-integer onsets, Float instead of Int (#200)
- Julia
Published by github-actions[bot] almost 2 years ago
Unfold - v0.7.0
Unfold v0.7.0
This is a bigger release 🎉 with some clearly breaking changes. For most users, those breaking changes should be near "invisible", but we can't promise it. Update as always without warranty.
Breaking:
- New major types:
- all
UnfoldModel-types are now parametric types with e.g. Float64 or Missing,Float64, which will make specializations in fitting much easier in the future (e.g. no missing values => faster fit!) - Internal: New Trait
ContinuousTimeTraitmaking it easier to specify custom types UnfoldModelnow takes a vector ofAbstractDesignMatrices, where eachAbstractDesignMatrixholds only one formula, one table and one modelmatrix. If you used accessorfunctions likeeventsandformulasnot much should change. Those accessor functions were harmonized to produce vectors of content.
- all
- FIR by default does not linearly interpolate anymore. This will help throughout the toolbox with minimal consequences on model-accuracy. That is, if an event-latency is in between two datasamples, it used to be linearly "split up". This now needs to manually activated using interpolate=true in firbasis. In the case of interpolate=false only integer event-latencies are allowed. In practice, this should have little impact, but it will change your numbers slightly.
Dict(:A=>(formula,times))from dict to vector of pairs[:A=>(formula,times]coeftablecolumnbasisnamerenamed toeventname, basisfunctions now need to have the samenameas the event in the Design-specification
Public renaming (breaking)
old | new
--|--
predict | predicttable (predict now returns a vector of arrays
formula| formulas (now returns a vector of StatsFormulas for all models)
Internal renaming
old | new
--|--
combineDesignmatrices | combine_designmatrices
equalizeLengths | equalize_lengths
zeropad | equalize_size
shiftOnset | shift_onset
get_modelmatrix | modelmatrices
randomeffectgroupings| random_effect_groupings
changeMatSize! | change_modelmatrix_size!
equalizeReMatlengths! | equalize_ReMat_lengths!
changeReMatSize! | change_ReMat_size!
lmm_combineMats! | lmm_combine_modelmatrix!
extract_symbol | termname
dropMissingEpcohs| drop_missing_epochs
DesignMatrix.Xs| DesignMatrix.modelmatrix
New functions & features:
- BasisFunction now has a method height, in addition to it's StatsModels.width
- pretty Term.jl output for models and designmatrices
- massunivariate models by default calculate SE now
- many more docstrings
- documentation fixes @vladdez
- residual(uf,data) function
- fit/solver show_progress controls output
Closes: #192 #161 #52 #189 #155
Merged pull requests: - readability - index.md (#187) (@vladdez) - readability - tutorials/howto (#188) (@vladdez) - readability - howto/lmm_pvalues (#191) (@vladdez) - unfold breaking 0.7 (#192) (@behinger)
Closed issues: - Documentation improvement suggestions (#189) - FileIO.load does not work (#194)
- Julia
Published by github-actions[bot] almost 2 years ago
Unfold - v0.6.10
Unfold v0.6.10
speed up effects load older JLD2 files
Merged pull requests: - backward compatible loading of JLD2 files (#176) (@behinger) - betterBackwardJLD2 (#177) (@behinger) - Update effects.jl (#179) (@behinger) - docs: add vladdez as a contributor for bug (#180) (@allcontributors[bot]) - Update basisfunctions.md - readability (#184) (@vladdez)
Closed issues: - Confusing documentation about Effects (#178) - Confusing usage of acronyms (#181) - firbasis shows error when name is not specified (#183) - Better error (message) when swapping basisfunction and formula in the event dict tuple (#186)
- Julia
Published by github-actions[bot] almost 2 years ago
Unfold - v0.6.9
Unfold v0.6.9
- stderror was wrongly forced to type of data, but should be type of coefficients. This lead to very rare problem (introduced in 0.6.6) thanks @vladdez for reporting
- LMMs had a major bug #175. If you had a model of type
0~ [...] + (1|item) + (1+A|subject)where the first term had fewer random slopes than the second, then thecoeftable(model)had wrongly orderedcoefnamesfor the random effects. This is fixed now - fixed LMM bugs in the tutorial (following from UnfoldSim major v0.4).
Merged pull requests: - Fix LMM, Fix STDerror bug (#174) (@behinger) - Fix LMM Reorder bug (#175) (@behinger)
- Julia
Published by github-actions[bot] about 2 years ago
Unfold - v0.6.5
Unfold v0.6.5
- JuliaFormatter on everything
- added
EffectsforUnfoldLinearMixedModels - updated docs & docstrings
Merged pull requests: - Create darus-dataverse.yml (#158) (@behinger) - UnfoldMakie0.4 (#159) (@behinger) - effects for LMMs (#160) (@behinger) - Update CITATION.cff (#162) (@jschepers) - fix for SE calculation breaking CI (#163) (@behinger) - fix #124, assert min number of spines; compat 0.17 BSplineKit (#164) (@behinger) - added docstrings for types fix #133 (#165) (@behinger) - JuliaFormatter (#166) (@behinger)
Closed issues: - 2nd level tools / plots / stats (#13) - [docs] Timebasis is empty? (#102) - Krylov + multithreading crashed (#122) - add errormessage if n_spl = 3 is used (#124) - Missing doc strings and types & functions in the Unfold documentation (#133)
- Julia
Published by github-actions[bot] about 2 years ago
Unfold - v0.6.4
Unfold v0.6.4
- Fix bug with multi-channel mixed models extension reshaping
- greatly enhanced documentation
Merged pull requests: - Update julia_call HowTo (#132) (@jschepers) - Updated "Adding documentation" section in the README (#134) (@jschepers) - Update README.md (#135) (@jschepers) - Update README.md (#136) (@jschepers) - added the @all-contributors functionality to the toolbox and added the emoji key (#142) (@ReboreExplore) - docs: add ReboreExplore as a contributor for infra (#143) (@allcontributors[bot]) - docs: add jschepers as a contributor for bug, code, and 5 more (#144) (@allcontributors[bot]) - docs: add behinger as a contributor for bug, code, and 9 more (#145) (@allcontributors[bot]) - docs: add ReneSkukies as a contributor for bug, and doc (#146) (@allcontributors[bot]) - docs: add palday as a contributor for code, and infra (#147) (@allcontributors[bot]) - docs: add kleinschmidt as a contributor for doc (#148) (@allcontributors[bot]) - docs: add ssaket as a contributor for bug (#149) (@allcontributors[bot]) - docs: add suddha-bpn as a contributor for bug (#150) (@allcontributors[bot]) - Add window length explanation to Docs (#151) (@ReneSkukies) - How to save and load unfold models (#152) (@jschepers) - docs: add ReneSkukies as a contributor for tutorial (#153) (@allcontributors[bot]) - update autoformat test (#156) (@behinger)
Closed issues: - Missing information about Extensions? (#138) - Outdated information in "Traditional Mass Univariate Analysis" tutorial (#139)
- Julia
Published by github-actions[bot] about 2 years ago
Unfold - v0.6.2
Unfold v0.6.2
- added rePCA for LMMs
- some minor bugfixes
- greatly improved type-stability, especially for time_expand
- potentially breaking: starting to deprecate FIRbasis.kernel field. kernel(firbasis) does not return the kernelfunction anylonger, which was required to be an anonymous function. Rather, it is replaced by kernel(firbasis,e) to directly calculate the kernel at e. While the FIRBasis Type changed, we have a @deprecate in place that should catch the old access.
Merged pull requests: - faster timeexpand (#127) (@behinger)
- Julia
Published by github-actions[bot] over 2 years ago
Unfold - v0.6.1
Unfold v0.6.1
- generate compat with breaking StatsModels v0.7 (@behinger)
- much improved speed of designmatrix generation in case of special circumstances of same event-kernel length for all events (the typical case) (@behinger)
- minor bugfixes
Merged pull requests: - improved column generation for designmatrices (#123) (@behinger) - StatsModels 0.7 compat + bugfixe (#125) (@behinger)
- Julia
Published by github-actions[bot] over 2 years ago
Unfold - v0.6.0
Unfold v0.6.0
Breaking
- Introducing PackageExtensions #119:
- In order to use MixedModels you have to manually load using MixedModels
- In order to use spl(ABC,5) splines, you have to manually load using BSplineKit
- In order to use solver_robust you have to manually load using RobustModels
This features reduced the package loading time from 15-25s to just 2s on my computer 🚀.
New Features
- massive ⚡ speed-ups⚡ using the following "tricks":
- New extension: GPU support (via using Krylov and solver_krylov(...;gpu=true), currently only CUDA support, but this is not a fundamental limitation, merely an implementation wise, 75s => 3s speedup! together with @jschepers
- Multithreading (which is the new default), speedup 2-10x
- Up to 80% smaller designmatrices (and thus faster fits), by removing unnecessary zeros via dropzeros - speedup untested
- ProgressBar in all solvers (can be deactivated)
Bugfixes - fix bug with LMM random effects wrongly reordered #115,#33 - 2D data reshaping not working correctly #110, #83 - replaced some unnecessary dependencies (e.g. DSP)
Closed issues: - Random effects are reordered (#33) - Regularization, Feature Importance for back-to-back regression (#74) - Massunivariate LMM with two dimensions results in "coef" error (#83) - bug: LMM mass univariate with 2-dim data (#110) - LMMs: coeftable re-order bug (#115) - Weak Dependencies (#119)
Merged pull requests: - bugfix reshape + LMM (#117) (@behinger) - pkgExtensions (#120) (@behinger) - krylov, GPU, progressbar, multithreading (#121) (@behinger)
- Julia
Published by github-actions[bot] over 2 years ago
Unfold - v0.5.1
Unfold v0.5.1
Closed issues: - Documentation: only dev available (#73) - What's up with LRT? (#101) - [docs] custom solvers (#103)
Bugfix release. - fix issues in documentation (update to new major UnfoldMakie version, upgrades some examples to use UnfoldSim) - fix issue with missings in Effects.jl (occuring in out-of-domain prediction with splines)
- Julia
Published by github-actions[bot] over 2 years ago
Unfold - v0.3.13
Unfold v0.3.13
Closed issues: - renaming things / refactoring / modifying output (#41) - Safe Predict for timeexpanded splines (#49) - Wrong dispatch with Any=>(f,times) and 2D-Data (#51) - Predict does not automatically get times (#53) - Doc: takes 3h (??) (#86) - Documentation: Explanation/ HowTo on Effects both lead to 404 page (#91) - WARNING: could not import Effects.referencegrid into Unfold (#93)
Merged pull requests: - improved tests for fit (#95) (@behinger) - first try to fix docs, 2h baseline (#96) (@behinger) - bugfixes (#97) (@behinger) - Remove unnecessary packages (#100) (@behinger)
- Julia
Published by github-actions[bot] over 3 years ago
Unfold - v0.3.12
Unfold v0.3.12
Closed issues: - do we need times(BasisFunction) and colnames(BasisFunction) or can they be the same? (#55)
Merged pull requests: - improved spline interface (#88) (@behinger) - effects - expand_grid (#94) (@vladdez)
- Julia
Published by github-actions[bot] over 3 years ago
Unfold - v0.3.11
Unfold v0.3.11
fixes bug #76, now more than two basisfunctions are actually correctly possible (previously only two "events" were saved, now it is correctly appended)
Closed issues: - effects with three basis functions (#76) - Rounding error in Unfold.epoch (#78)
Merged pull requests: - rounding latency .5 fixes #78 (#80) (@behinger)
- Julia
Published by github-actions[bot] almost 4 years ago
Unfold - v0.3.8
Unfold v0.3.8
- effects function to easily get marginal means (think emmeans) or "adjusted ERPs"
- likelihoodratiotest for mass-univariate LMMs
- greatly improved documentation (including start of shift to literate.jl)
- speed improvements due to implace solving; up to 2x speed-up compared to matlab
- several bugfixes regarding coeftable
Closed issues: - bug: predict + two basisfunctions (#60) - bug: minfo is currently not saved (#63) - coeftable output for MM with two random effects seems off (#65)
Merged pull requests: - activate benchmark again (#57) (@behinger) - Inplace Solver (#64) (@behinger) - LMM LRtest (#67) (@behinger) - LMM+timeexpansion coeftable output broken (#68) (@behinger) - Fix Documentation (#69) (@behinger) - fix bug with typify and single basis function (#71) (@behinger) - literate (#72) (@behinger)
- Julia
Published by github-actions[bot] about 4 years ago
Unfold - v0.3.5
Unfold v0.3.5
Closed issues: - naming of output (#20)
Merged pull requests: - Effects Interface (#50) (@behinger) - UnfoldMakie is a package (#54) (@behinger) - Pull to finally fix the benchmark... (#56) (@behinger) - v0.3.5 (#58) (@behinger)
- Julia
Published by github-actions[bot] about 4 years ago
Unfold - v0.3.1
Unfold v0.3.1
Closed issues:
- Fix Documentation (#34)
- typo in fit.jl test (#38)
- LoadError: Lines not defined while using plot.jl (#39)
Merged pull requests: - Corrected standard error calculations (#37) (@Zebrakopf) - add research notice (#40) (@yochannah) - Restructure (#45) (@behinger) - BasisFunction restructure (#46) (@behinger) - fix Project.toml (#47) (@palday)
- Julia
Published by github-actions[bot] over 4 years ago