Skip to content

Latest commit

 

History

History
313 lines (224 loc) · 9.81 KB

NEWS.rst

File metadata and controls

313 lines (224 loc) · 9.81 KB
Raw

BrainIAK 0.10 (2020-02-06)

Features

  • reconstruct: Add inverted encoding model (IEM) for recreating continuous representations in new package for reconstruction methods. (#364)
  • funcalign: Add FastSRM module for input data that does not fit in RAM. (#421)
  • fmrisim: Update to how drift is calculated and used. (#437)
  • searchlight: Allow processing single voxels. (#440)
  • eventseg: Add split-merge feature to improve fits when event lengths are highly uneven. (#451)

Documentation improvements

  • Add --no-use-pep517 Pip installation flag for compatibility with Pip >= 19.1.1. (#454)

BrainIAK 0.9.1 (2019-08-27)

Features

  • Added Conda packages for Python 3.7.

BrainIAK 0.9 (2019-06-18)

Features

  • isc: Revamped ISC/ISFC functionality with more statistical tests. (#384)
  • utils: phase_randomize outputs phase-scrambled input data, not tied to ISC/ISFC. (#396)
  • utils: p_from_null no longer estimates distribution, simply returns p-value. (#397)
  • isc: Analyses will now tolerate NaNs or tolerate a proportion of NaNs. (#398)
  • isc: Output either vectorized triangle and diagonal or square matrices. (#399)
  • isc: Asymmetric ISFC analysis can now be performed based on a targets array. (#409)
  • Added tutorials to Docker image. (#425)

Deprecations and removals

  • isfc: ISC/ISFC have a new API and are available in brainiak.isc, not brainiak.isfc. (#384)

BrainIAK 0.8 (2018-11-06)

Features

  • searchlight: Add ball shape. (#348)
  • eventseg: Added event_chains option and model_prior function. (#353)
  • srm: Added the ability to transform a new subject to a shared response. (#371).
  • fmrisim: Added fitting procedure for SFNR and SNR parameters. Updated AR to be ARMA, involving both the generation and estimation. (#372)
  • searchlight: Halved memory usage when pool_size=1. (#386)
  • reprsimil: Added an option in BRSA to set the prior of SNR to be "equal" across all voxels. (#387)

Bugfixes

  • fmrisim: Corrected error in generating system noise; specifically spatial noise was being double counted. Updated the export epoch file to deal with error in epoch number generation. (#372).
  • utils: Fix AFNI-style design matrix generation. (#381)

Documentation improvements

  • fcma: Clarify that image data is shuffled. (#365)
  • Fix Conda channels in install instructions. (#373)

Deprecations and removals

  • fmrisim: Removed plot_brain because other tools like nilearn do a much better job at plotting data. (#372)

BrainIAK 0.7.1 (2018-02-20)

Features

  • reprsimil: Changed the default optimizer of (G)BRSA to L-BFGD-B. (#337)

Bugfixes

  • eventseg: Fixed bug that was causing fits to be asymmetric (#339)

Documentation improvements

  • reprsimil: Added explanation that both BRSA and GBRSA assume zero-mean in the distribution of beta patterns. (#337)

BrainIAK 0.7 (2018-02-12)

Features

  • funcalign: Added the Robust Shared Response Model method. (#302)
  • fmrisim: Update convolution and drift. (#309)
  • eventseg: Added option to compute p values for ISC and ISFC (#310)
  • Added Conda packages. (#328)

Documentation improvements

  • Updated the searchlight API docs. (#324)

BrainIAK 0.6 (2017-11-10)

Features

  • reprsimil: Add Group Bayesian RSA, add transform (decoding) and score (model selection) functions, add automatic determination of the number of necessary nuisance regressors. utils: modified gen_design to make the generated design matrix approximately scaled in amplitudes. (#194)
  • searchlight: Improved performance via tweaked multiprocessing usage. (#240)
  • fmrisim: Updated drift calculation and masking. (#244)
  • eventseg: Add set_event_patterns() and notebook example (#266)
  • Added a Docker image for testing BrainIAK without installing. (#272)
  • eventseg: Fixed numerical instability bug, added utility function for weighted variance (#292)
  • Restricted multiprocessing to the available number of CPUs. (#293)

Bugfixes

  • searchlight: Do not use sys.exit. (#156)
  • reprsimil: Follow random number guidelines. (#239)

Deprecations and removals

  • Changed GitHub organization to BrainIAK. Update your remote URLs. (#277)

BrainIAK 0.5 (2017-05-23)

Features

  • FCMA partial similarity matrix option. (#168)
  • Faster FCMA cross validation via multiprocessing. (#176)
  • Inter-subject correlation (ISC) and inter-subject functional correlation (ISFC). (#183)
  • Input/output and image modules with high-level APIs. (#209)
  • FCMA support for random permutations. (#217)
  • A distributed version of SRM. (#220)
  • Shape masks for the searchlight. (#221)

Deprecations and removals

  • Changed fmrisim to compute signal-to-fluctuation-noise ratio (SFNR) instead of signal-to-noise ratio (SNR). (#224)

BrainIAK 0.4 (2017-01-19)

Features

  • Distributed searchlight. (#148)
  • Multi-voxel pattern analysis (MVPA) support in FCMA. (#154, #157)
  • Fast Pearson correlation coefficient computation. (#159)

BrainIAK 0.3.2 (2016-10-31)

Features

  • Faster event segmentation via Cython implementation. (#111)
  • fMRI data simulator (fmrisim). (#135)

BrainIAK 0.3.1 (2016-09-30)

Features

  • Event segmentation. (#72)
  • Full correlation matrix analysis (FCMA). (#97, #122)
  • Bayesian representational similarity analysis (BRSA). (#98)
  • Deterministic SRM. (#102)
  • Semi-supervised shared response model (SSSRM). (#108)

BrainIAK 0.3 (2016-09-30) [YANKED]

BrainIAK 0.2 (2016-08-03)

Features

  • Hyperparameter optimization. (#58)

Deprecations and removals

  • Removed _ from package names. (#73)

BrainIAK 0.1 (2016-07-12)

Features

  • Initial release, including:
    • Shared response model (SRM).
    • Topographic factor analysis (TFA) and hierarchical topographical factor analysis (HTFA).