NIFTy merge requestshttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests2018-03-15T17:04:25Zhttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/225Feature/operator overloading sugar2018-03-15T17:04:25ZSilvan StreitFeature/operator overloading sugarThis could simplify the notation by adding a nice way to write the inverse of an operator as simply `~D`.This could simplify the notation by adding a nice way to write the inverse of an operator as simply `~D`.https://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/409Simplify licensing2020-05-13T11:36:53ZGordian EdenhoferSimplify licensingFixes #283 .Fixes #283 .https://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/867fixed minor typo2023-09-14T20:05:54ZDavid Gorbunovfixed minor typopirmal -> primal @gedenhofpirmal -> primal @gedenhofhttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/865Draft: Enable likelihoods to operate over pytrees2023-09-07T15:05:04ZPhilipp FrankDraft: Enable likelihoods to operate over pytreesTree-maps and Tree-reduces the likelihood functions to operate on arbitrary tree-like structures.Tree-maps and Tree-reduces the likelihood functions to operate on arbitrary tree-like structures.Philipp FrankPhilipp Frankhttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/864Tree-map poissonian2023-09-07T11:57:19ZPhilipp FrankTree-map poissonianPhilipp FrankPhilipp Frankhttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/861Draft: JaxOpt style optimization for jifty VI2023-09-19T14:46:33ZPhilipp FrankDraft: JaxOpt style optimization for jifty VIImplements a JaxOpt style optimization for jifty VI inference: `OptimizeVI`
Additionally reimplements `optimize_kl` using it.Implements a JaxOpt style optimization for jifty VI inference: `OptimizeVI`
Additionally reimplements `optimize_kl` using it.Philipp FrankPhilipp Frankhttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/858Gaussian lh for complex data2023-07-31T15:23:11ZJakob RothGaussian lh for complex dataAs done in VC Gaussian, use vdot to compute the sum of squared residuals.
@gedenhof Shouldn't there be a cast to real numbers? By just using vdot the result is still complex with negligible imaginary part.As done in VC Gaussian, use vdot to compute the sum of squared residuals.
@gedenhof Shouldn't there be a cast to real numbers? By just using vdot the result is still complex with negligible imaginary part.https://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/853Fix nifty.re "sample_evi" docstring2023-06-05T20:53:21ZJakob RothFix nifty.re "sample_evi" docstringfix "sample_evi" docstring in "kl.py".fix "sample_evi" docstring in "kl.py".Gordian EdenhoferGordian Edenhoferhttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/850Rethink sampling2023-07-04T16:45:50ZGordian EdenhoferRethink samplingGive more control over how to map the drawing of samples to the user. In doing so restructure the way minimization works.Give more control over how to map the drawing of samples to the user. In doing so restructure the way minimization works.Gordian EdenhoferGordian Edenhoferhttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/844Fix the NIFTy pipeline2023-03-17T17:00:40ZGordian EdenhoferFix the NIFTy pipelineGordian EdenhoferGordian Edenhoferhttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/841Test less2023-03-17T16:58:11ZGordian EdenhoferTest lessPhilipp FrankPhilipp Frankhttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/805Use metric sample for probing2022-11-01T11:39:40ZPhilipp FrankUse metric sample for probingApproximates tr(log(M)) using tr(log(T)) where T is the projection of M into the krylov subspace K(M, v) where v is a sample from the metric M (i.E. v = v_lh + v_pr where v_lh/v_pr are samples from the likelihood/prior metric, respective...Approximates tr(log(M)) using tr(log(T)) where T is the projection of M into the krylov subspace K(M, v) where v is a sample from the metric M (i.E. v = v_lh + v_pr where v_lh/v_pr are samples from the likelihood/prior metric, respectively. In addition, the projected sample is constructed by taking v_pr projecting out the subspace K(M,v) using its eigen-basis. This ensures that both, the prior dominated part of v and the part already covered by tr(log(T)) is projected out.Philipp FrankPhilipp Frankhttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/803Implement geoMAP2022-11-27T15:19:58ZGordian EdenhoferImplement geoMAPUnfortunately, it performs much poorer than expected.Unfortunately, it performs much poorer than expected.Gordian EdenhoferGordian Edenhoferhttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/802Transition function2022-10-28T15:13:31ZMargret WesterkampTransition functionChange of transition operator to transition function as well as change of the corresponding testsChange of transition operator to transition function as well as change of the corresponding testsPhilipp Arrasparras@mpa-garching.mpg.dePhilipp Arrasparras@mpa-garching.mpg.dehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/801Config multifrequency trans2022-10-25T12:27:43ZMargret WesterkampConfig multifrequency transPhilipp Arrasparras@mpa-garching.mpg.dePhilipp Arrasparras@mpa-garching.mpg.dehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/800Config multifrequency trans2022-10-24T14:04:42ZMargret WesterkampConfig multifrequency transPhilipp Arrasparras@mpa-garching.mpg.dePhilipp Arrasparras@mpa-garching.mpg.dehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/799Config multifrequency trans2022-10-24T13:50:57ZMargret WesterkampConfig multifrequency transPhilipp Arrasparras@mpa-garching.mpg.dePhilipp Arrasparras@mpa-garching.mpg.dehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/798Config multifrequency trans2022-10-24T12:48:47ZMargret WesterkampConfig multifrequency transPhilipp Arrasparras@mpa-garching.mpg.dePhilipp Arrasparras@mpa-garching.mpg.dehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/797Config multifrequency trans2022-10-24T11:33:21ZMargret WesterkampConfig multifrequency transPhilipp Arrasparras@mpa-garching.mpg.dePhilipp Arrasparras@mpa-garching.mpg.dehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/793Draft: Misc edh2023-01-13T11:37:39ZPhilipp Arrasparras@mpa-garching.mpg.deDraft: Misc edhGordian EdenhoferGordian Edenhofer