Skip to content

Releases: acerbilab/bads

v1.1.3

05 Dec 17:27
74919c0

Choose a tag to compare

Bug Fix

  • Fixed typo in private/gpupdate.m:262: error_indexerr_index (undefined variable causing errors when objective function returns NaN/Inf values)

Maintenance

  • Added scripts/update_version.sh for easier version management
  • Added CLAUDE.md with release instructions

Fixes #17

v1.1.2

14 Nov 15:20
019f0b4

Choose a tag to compare

v1.1.2

  • Return ysd_vec (estimated standard deviations of final samples) in output for user-specified noise.
  • Compute mean and std of the final estimate using fsd of each sample for user-specified noise.
  • Revert Ninit default to nvars.
  • If x0 is not provided, sample uniformly randomly from the plausible box (instead of selecting the middle of the plausible box).

v1.1.1

31 Oct 15:33
a21f2ee

Choose a tag to compare

v1.1.1

  • Full implementation of user-provided noise and support for heteroskedastic noise.
  • Several updates to vbmc_examples.m, and added user-specified noise example (replacing former Example 4).
  • Several changes to README.md.
  • Fixed bug of initial GP mean function value (used to be 0).
  • Mild refactoring; deleted some unused code.
  • Updated OPTIONS.Ninit default to 10 + nvars.

v1.0.8

09 May 12:07
ba31706

Choose a tag to compare

Additional fixes to uncertainty handling and user-specified noise and their internal options; added versioning; always print target type (deterministic or stochastic).

v1.0.7

06 May 07:02
b83c2c0

Choose a tag to compare

Fixed bug with user-specified noise when removing points in the Gaussian process training (user-specified noise values from gpstruct.s were not removed, causing training to fail).

v1.0.6

22 Mar 19:52
ded3fa7

Choose a tag to compare

Added support for specified target noise; fixed tic/toc issues; minor fixes.