Changelog¶
2024-02-14 v0.9.7¶
Updata data_handling to comply with pandas 3.0
Consider declination periodicity in residual calculation
2023-12-11 v0.9.6¶
Add a consistency check for out of range latitudes and longitudes
Rewrite periodicity handling for declinations
2023-09-29 v0.9.5¶
Fix truncation errors exploding due to 1 / cos
2023-09-20 v0.9.4¶
Add function to calculate residuals
2023-09-19 v0.9.3¶
legacy attribute for PaleoKalmag model class
2023-09-13 v0.9.2¶
Fix using slope
Fix stream input
2023-09-13 v0.9.1¶
Remove torch.set_default_dype from corefieldmodel
2023-09-12 v0.9.0¶
Further accelerate torch implementation
gradient and nez2dif functions are removed, as they are now included in the filter directly
2023-09-12 v0.8.0¶
This is a breaking change!
We move lots of stuff to torch for speed improvement and gradient availability.
rename “co-lat” to “colat”
2023-09-01 v0.7.0¶
Fix an error in the pytorch marginal likelihood implementation (detach)
2023-08-15 v0.6.0¶
Update pymagglobal dsh_basis call structure
Fix age errors in read_data
2023-08-15 v0.5.8¶
Move scaling to pymagglobal
2023-05-24 v0.5.7¶
Add raw temporal errors to data
Set minimal age uncertainty of 10 years
2023-03-08 v0.5.6¶
Update errors when sampling a smoothed model
2022-12-02 v0.5.5¶
Add optional progressbar to dlib_wrap
Minor performance gain by circumventing another diagonal matrix product
2022-11-30 v0.5.4¶
Realize that the secular variation diagonal is constant and drop calculation from NIGP
2022-11-30 v0.5.3¶
Slight performance gain by circumventing a diagonal matrix
2022-11-29 v0.5.2¶
Fix an error with logdet calculation
2022-11-23 v0.5.1¶
Extend _calc_filehash to accept streams
2022-11-21 v0.5.0¶
Update outlier rejection and read_data: We now calculate a filehash and save a unique record ID, in order to reliably relate outlier rejection lists to their original data file
2022-10-11 v0.4.7¶
Add sampling from prior
2022-10-11 v0.4.6¶
Add progressbar to outlier removal
2022-09-16 v0.4.5¶
Fix the docstring of CoreFieldModel.reject_outliers and add smoothed property
2022-04-20 v0.4.4¶
Add plotting function paleokalmag.MapLocPicker for local analysis of datasets and models
2022-04-19 v0.4.3¶
address numerical instability in torch marginal likelihood
consider declination mapping
2022-04-05 v0.4.2¶
bugfixes in marglog
2022-04-01 v0.4.1¶
fix matrix inversion in torch marglog
2022-04-01 v0.4.0¶
provide a torch-accelerated version of the marginal likelihood
2022-03-18 v0.3.1¶
Parallelize for-loop in dsh_basis
2022-02-09 v0.3.0¶
Update dlib_wrap. The default is now exhausting the function calls
Add a check for args kwarg in dlib_opt
Add unittests for the new behaviour
2022-02-09 v0.2.1¶
Consider periodicity of declination
2022-02-03 v0.2.0¶
Call nez2dif only once in correction step
Fix docstring for reject_outliers
Add __len__ to data and fix __repr__
Add optional progress feedback for forward and backward step
2022-01-18 v0.1.3¶
Add a check for Fortran order and fix dsh_basis to only consider the spatial part
2022-01-17 v0.1.2¶
Change Matrix inversion to eigenvalue decomposition for stability issues
Minor performance gain by moving the calculation of likelihood terms
2022-01-10 v0.1.1¶
Windows installation fix
2022-01-07 v0.1.0¶
Change model name to PaleoKalmag
Enable random behavior in dlib and set as a default
Fix documentation typo
2021-12-08 v0.0.12¶
Minor bug fix
2021-11-19 v0.0.11¶
Add sampling to kalman
2021-11-11 v0.0.10¶
Update read_data and drop records with NaNs in location
2021-01-11 v0.0.9¶
Address issue #2 and move the spanning of the diagonal matrix out of the data routine.
2021-28-10 v0.0.8¶
Minor bug fixes
2021-20-10 v0.0.7¶
Move dt determination from Data to read_data
Document the input keys for Data
2021-26-08 v0.0.6¶
Monkey patch the warnings
2021-07-08 v0.0.5¶
add a test for dsh_basis with scaling
fix scaling behaviour of c++ version of dsh_basis
2021-07-06 v0.0.4¶
fix a bug in reading rejection lists
added a test for saving and loading
minor bug fixes
update example for outlier removal
2021-06-24 v0.0.3¶
pole safe version of c++ implementation of basis functions
polished the examples
2021-06-16 v0.0.2¶
changed the interface of
CoreFieldModelto work better with marglog resultsmarginal likelihood and optimization added
outlier rejection routine added
examples for marginal likelihood and rejection added
minor bug-fixes