NIFTy merge requestshttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests2021-10-07T08:10:09Zhttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/688Sample lists2021-10-07T08:10:09ZPhilipp Arrasparras@mpa-garching.mpg.deSample lists- [x] Minisanity redesign @parras
- [x] Changelog entry: `KL.samples` is something different now, have introduced `SampleList`, Geo and MGVI unified. @pfrank
- [x] Implement save and load for `SampleList` with MPI @parras
- [x] Docu for ...- [x] Minisanity redesign @parras
- [x] Changelog entry: `KL.samples` is something different now, have introduced `SampleList`, Geo and MGVI unified. @pfrank
- [x] Implement save and load for `SampleList` with MPI @parras
- [x] Docu for `SampleList` @parras
- [x] Check `SampleList` in plotting etc in demos because `KL.samples` is something different now. @pfrank
- [x] Check years in copyright @parras
- [x] Rename `global_iterator` @parrasMartin ReineckeMartin Reineckehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/696SampleList improvements2021-10-12T09:49:34ZPhilipp Arrasparras@mpa-garching.mpg.deSampleList improvementsThis merge requests revamps the save/load logic of `ift.SampleList`. Now every sample is written to its own pickle file.
Additionally, exporting to hdf5 files has been added. Loading from hdf5 is not supported.This merge requests revamps the save/load logic of `ift.SampleList`. Now every sample is written to its own pickle file.
Additionally, exporting to hdf5 files has been added. Loading from hdf5 is not supported.https://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/699fix link to nifty publications2021-10-27T07:44:49ZJakob Rothfix link to nifty publicationsThe link to the NIFTy publication was outdatedThe link to the NIFTy publication was outdatedhttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/697Sampling dtypes in sumops2021-10-27T08:19:30ZPhilipp Arrasparras@mpa-garching.mpg.deSampling dtypes in sumopsMartin ReineckeMartin Reineckehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/700MultiField2Vector: Support complex dtype2021-10-28T11:55:01ZPhilipp Arrasparras@mpa-garching.mpg.deMultiField2Vector: Support complex dtypeMartin ReineckeMartin Reineckehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/703CI: Adopt to new jupyter interface2021-11-13T10:01:30ZPhilipp Arrasparras@mpa-garching.mpg.deCI: Adopt to new jupyter interfacehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/704CI: Adopt to new jupyter interface2021-11-13T10:41:31ZPhilipp Arrasparras@mpa-garching.mpg.deCI: Adopt to new jupyter interfacehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/693More user friendly demos2021-11-15T12:03:49ZGordian EdenhoferMore user friendly demosYield descriptive error messages if invalid input is passed to the PoissonianEnergy likelihood. Furthermore, never start the minimization from exactly zero as this is potentially unstable in general and users tend to copy-paste code from...Yield descriptive error messages if invalid input is passed to the PoissonianEnergy likelihood. Furthermore, never start the minimization from exactly zero as this is potentially unstable in general and users tend to copy-paste code from the demos for new project.
Starting close to or even worse exactly at zero with MGVI is error prone for simple models with a linear response and a simple correlated field. The main problem are the zeros that appear in the Jacobians for everything but the excitations as a result of the linearity of the model in exactly these. While for the minimization itself zero-curvature for the Fisher metric is not necessarily problematic because of the prior curvature, it results in problems for the sampling. Namely sampling along axis with zero Fisher curvature results in prior samples being drawn for these axis.
On the one hand this is desirable because in the absence of information from the likelihood, the prior should be the only quantity that matters. On the other hand, this results in the model being sometimes impossible to optimize appropriately. This is because often for a flexible model the prior samples for the non-linear parameters are too diverse to jointly optimize their mean in a sensible way. As an example, for prior samples for the fluctuations parameter, since samples with a too high fluctuation parameter compared to the data are compensated by samples with a too low fluctuation parameter compared to the data. However, since it is critical to fix the fluctuations parameter before anything else can be optimized in a sensible way, in essence no optimization can nor does happens.Martin ReineckeMartin Reineckehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/706setup.py: Shorten description2021-11-15T12:21:23ZPhilipp Arrasparras@mpa-garching.mpg.desetup.py: Shorten descriptionSince setuptools 59.1.0 multi-line descriptions are not allowed anymore.Since setuptools 59.1.0 multi-line descriptions are not allowed anymore.https://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/708Cut down CI time a little2021-11-15T13:39:24ZPhilipp Arrasparras@mpa-garching.mpg.deCut down CI time a littleI have made sure that the conceptional output of the changed demos has not changedI have made sure that the conceptional output of the changed demos has not changedMartin ReineckeMartin Reineckehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/707setup.py: Shorten description2021-11-15T13:39:47ZPhilipp Arrasparras@mpa-garching.mpg.desetup.py: Shorten descriptionSince setuptools 59.1.0 multi-line descriptions are not allowed anymore.Since setuptools 59.1.0 multi-line descriptions are not allowed anymore.https://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/705Fix performance regression2021-11-15T16:24:31ZPhilipp Arrasparras@mpa-garching.mpg.deFix performance regressionintroduced by 2db6c449d6d3d9b96c33473a05c00009c993a0c6introduced by 2db6c449d6d3d9b96c33473a05c00009c993a0c6https://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/711updated minimum Python version to 3.7 (3.6 EOL is Dec 2021)2021-11-16T12:47:34ZMartin Reineckeupdated minimum Python version to 3.7 (3.6 EOL is Dec 2021)Since Nifty 8 is probably not released before December, I guess we can apply this right now...Since Nifty 8 is probably not released before December, I guess we can apply this right now...https://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/710Various improvements2021-11-16T14:33:30ZPhilipp Arrasparras@mpa-garching.mpg.deVarious improvementsMartin ReineckeMartin Reineckehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/597Add file for simplified minimization2021-11-16T14:33:30ZPhilipp Arrasparras@mpa-garching.mpg.deAdd file for simplified minimization- [x] For likelihood, n_samples, newton_convergence, ... also support function that takes iteration index and returns the respective value (for dynamic optimization)
- [x] Add support for constants (also potentially variable during optim...- [x] For likelihood, n_samples, newton_convergence, ... also support function that takes iteration index and returns the respective value (for dynamic optimization)
- [x] Add support for constants (also potentially variable during optimization)
- [x] Latent: Do not save hdf5
- [x] 2d samples einzeln plotten (max. 20 oder so)
- [x] Figure out how to automatically visualize residuals
- [x] Add slot for ground truth
- [x] Add tests
- [x] Support constants and point estimates with domain errors
!710 is a part of this one. It may either be merged first or automatically be closed by merging this one.Martin ReineckeMartin Reineckehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/712Fix inverse of special distribution operators and add tests2021-11-22T14:11:31ZPhilipp Arrasparras@mpa-garching.mpg.deFix inverse of special distribution operators and add testsPhilipp Arrasparras@mpa-garching.mpg.dePhilipp Arrasparras@mpa-garching.mpg.dehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/713Fix inverse of special distribution operators and add tests2021-11-22T14:14:49ZPhilipp Arrasparras@mpa-garching.mpg.deFix inverse of special distribution operators and add testsPhilipp Arrasparras@mpa-garching.mpg.dePhilipp Arrasparras@mpa-garching.mpg.dehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/714Optimize kl evolution2021-11-24T09:32:46ZPhilipp Arrasparras@mpa-garching.mpg.deOptimize kl evolutionFixes #339Fixes #339Martin ReineckeMartin Reineckehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/716optimize_kl: sanity check of input and minor stuff2021-12-06T13:00:17ZPhilipp Arrasparras@mpa-garching.mpg.deoptimize_kl: sanity check of input and minor stuffMartin ReineckeMartin Reineckehttps://gitlab.mpcdf.mpg.de/ift/nifty/-/merge_requests/717Merge branch 'imports' into 'NIFTy_8'2021-12-06T13:17:50ZPhilipp Arrasparras@mpa-garching.mpg.deMerge branch 'imports' into 'NIFTy_8'Martin ReineckeMartin Reinecke