This C++ version of BAT is still being maintained, but addition of new features is unlikely. Check out our new incarnation, BAT.jl, the Bayesian analysis toolkit in Julia. In addition to Metropolis-Hastings sampling, BAT.jl supports Hamiltonian Monte Carlo (HMC) with automatic differentiation, automatic prior-based parameter space transformations, and much more. See the BAT.jl documentation.
Performance testing
BAT comes with a test suite which performs benchmark and performance tests. The test suite is run for each release and the results can be found here.
The test suite comprises of several types of tests which each come
with a set of sub-tests. So far, only tests for the MCMC are
implemented. Each test returns a status which is one of the following:
| good | the test is successful |
| acceptable | the test is successful but at least one sub-test result deviates from the expectation to a level which is still acceptable |
| bad | the test is not successful, at least one sub-test result deviates from the expectation to a level which is not acceptable |
| fatal | the test is not successful, something is seriously wrong |
| unknown | the outcome of the test is not known |
| off | test is not performed |
Each test first summarizes the settings used and then lists the sub-tests with the results and allowed deviations (tolerances). The tests implemented so far are:
- the auto-correlation for each parameter,
- the chi2-value,
- the Kolmovorov-Smirnov probability,
- the mean value,
- the mode,
- the variance,
- the 10% to 90% quantiles.
- the auto-correlation for each parameter,
- the chi2-value.