Acerbilab Vbmc Versions Save

Variational Bayesian Monte Carlo (VBMC) algorithm for posterior and model inference in MATLAB

v1.0.12

1 year ago

v1.0.12 is a quick-fix to v1.0.11:

  • Added for each prior function the log-pdf function as well (numerically better).
  • Fixed Example 5 (and minorly 6) to also use the new log pdfs.

v1.0.11

1 year ago

Updated to v1.0.11:

  • Added several new functions in shared folder for evaluating the pdf and sampling from prior distributions for both bounded and unbounded variables (uniform, trapezoidal, smoothed trapezoidal, smoothbox).
  • New example in vbmc_examples.m describing usage of these priors.
  • Added test folder for unit tests.
  • Minor bug fix in the initialization of variational posteriors (in vbinit_vbmc.m).
  • Added bibtex references to the readme.

v1.0.10

1 year ago

Minor speed improvements to the VIQR acquisition function (default for noisy likelihoods).

v1.0.9

1 year ago

Minor fixes related to the bounded transforms.

v1.0.8

2 years ago

Minor fix: Fixed uuinv in fminfill.

v1.0.7

2 years ago
  • Bug fix: Corrected gradient computation for the variational weights in the Monte Carlo entropy (thanks to Chengkun Li).

v1.0.6

2 years ago
  • code cleanup: removed some unused or experimental features, added comments.

v1.0.5

2 years ago
  • Fixed bug in slice sampler that slightly biased sampling near the provided bounds (used in GP hyperparameter sampling).

v1.0.4

3 years ago
  • Modified VBMC output to return SAMPLES struct (might be used in the future).
  • Fixed starting points within bounds in vbmc_mode.m

v1.0.3

3 years ago

Fixed bug and added unit tests for D=1; updated options function.