Bibliographic record
Abstract
Most work on NiBabel so far has been by Matthew Brett (MB), Chris Markiewicz (CM), Michael Hanke (MH), Marc-Alexandre Côté (MC), Ben Cipollini (BC), Paul McCarthy (PM), Chris Cheng (CC), Yaroslav Halchenko (YOH), Satra Ghosh (SG), Eric Larson (EL), Demian Wassermann, Stephan Gerhard and Ross Markello (RM). References like "pr/298" refer to github pull request numbers. 4.0.0 (Saturday 18 June 2022) New feature release in the 4.0.x series. New features Add 'mask', 'compat' and 'smallest' dtype aliases to NIfTI images to allow for dtype specifications that can depend on the contents of the data. 'mask' is a synonym for uint8. 'compat' will find the nearest Analyze-compatible (therefore widely supported) dtype that will not truncate the data. 'smallest' attempts to find the smallest integer dtype that will contain the data. (pr/1096) (CM, reviewed by Chris Rorden and Josh Teves) Add dtype arguments to Cifti2Image (pr/1111) (CM) Allow dtypes to be passed to Analyze-like images at __init__() and to_filename() to provide better control over output images. (pr/1082) (CM, following discussions with Chris Rorden, Josh Teves, Jerome Dockes, and MB) Allow compressed GIFTI images (MB, reviewed by CM) Add zstd compression support (pr/1005) (Andrew Van, reviewed by CM) Support ExternalFileBinary GIFTI data arrays (PM, reviewed by CM) Enhancements Document InTemporaryDirectory as non-thread-safe (pr/1103) (Jacob Roberts, reviewed by MB) Unify Caret-XML-style metadata structure (GiftiMetaData, Cifti2MetaData) as dict-like (pr/1091) (CM, reviewed by Josh Teves and Hao-Ting Wang) Add __repr__ methods to GIFTI objects (pr/1092) (CM, reviewed by Josh Teves and Hao-Ting Wang) Create gzip header deterministically by default (pr/1024) (CM, reviewed by YOH) Provide clear error message when files with zip extensions don't match file contents (pr/1013) (Jérôme Dockès, reviewed by CM) Bug fixes Re-import externals/netcdf.py from scipy to resolve numpy API change (pr/1110) (CM) Resize ArraySequence.data without helper function to avoid reference increment (pr/1093) (MC, reviewed by CM) Maintenance Update submodule URLs to use https over git protocol (pr/1097) (CM) Published BIAP 9: CoordinateImage API (pr/1084) (CM) Drop uses of deprecated distutils (pr/1073) (CM, reviewed by MB) Suppress LGTM false alarm "Clear-text logging of sensitive information" (pr/1052) (Dimitri Papadopoulos, reviewed by CM) Test on Python 3.10 (pr/1047) (CM) Fix typos found by codespell (pr/1040, pr/1044) (Dimitri Papadopoulos, reviewed by CM) Run stable tests weekly, pre-release tests nightly (pr/1025) (CM) Documentation updates to establish/clarify governance and decision making (pr/1019, pr/1020, pr/1022, pr/1018, pr/1017, pr/1016) (MB and CM) API changes and deprecations Writing NIfTIs with 64-bit integer dtypes is getting harder. Passing (u)int64 arrays to Nifti1Image and subclasses will warn unless a header or dtype option is passed; in the future this will become an error. Additionally, passing int or 'int' to set_data_dtype() now raises an error, requiring an explicit numpy dtype to make 64-bit integer images. (pr/1082) (CM, following discussions with Chris Rorden, Josh Teves, Jerome Dockes, and MB) Drop support for Python 3.6, Numpy < 1.17 (pr/1079) (CM) Fully removed the following APIs, which have raised errors on use since 3.0 (pr/980) (CM, reviewed by Jonathan Daniel) nibabel.trackvis nibabel.volumeutils.calculate_scale nibabel.volumeutils.can_cast nibabel.volumeutils.scale_min_max nibabel.dataobj_images.DataobjImage.get_shape nibabel.minc1.MincImage (use Minc1Image) nibabel.minc1.MincFile (use Minc1File) nibabel.filebasedimages.FileBasedImage.from_files nibabel.filebasedimages.FileBasedImage.filespec_to_files nibabel.filebasedimages.FileBasedImage.to_filespec nibabel.filebasedimages.FileBasedImage.to_files nibabel.arrayproxy.ArrayProxy.header keep_file_open=="auto" parameter to load method (now must be boolean)
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.003 | 0.001 |
| Scholarly communication | 0.006 | 0.008 |
| Open science | 0.003 | 0.004 |
| Research integrity | 0.002 | 0.004 |
| Insufficient payload (model declined to judge) | 0.428 | 0.404 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; the direct Gemma label and the distilled Codex classifier agree on what is shown here.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".