Major update - CombinDS can now run on CUDA as recursion is replaced by iteration.
- Keep the old version of CombineDS that uses recursion just in case.
- Ignore vscode hidden directories.
- simBranch now avoids recursion
- replaces goesUndetected recursion with iteration
- Added posterior sampling
- Adapted to new rootppl build system
- Improved Python program for posterior analysis includes median, etc.
- New framework for running, which supports better statistics and cluster features.
- Newer version of logz plot.
- Safety improvements to avoid very rare assert (v > 0) error
- Minimum number must be hardcoded, library function application not possible in CUDA.
- Toolchain improvements.
- Fixed bugs in the GBM model
- Geometric Brownian motion added.
- This versions works with the new distributions from
phyrppl
after LinearNormalInverseGammaNormal was implemented.
Various small changes, mostly improvements to the script tools.
- can handle segfaults with runppl.sh
- improved final sampling
- some tests that lead to discovery of a bug in NomalInverseGammaNormal sampler
- posterior plotting in R improved
- runner script
- Major bug found, which affected AnaDS1 (not being executed).
- AnaDS-GBM implemented
- Improvements to tools.
Introduced templates to handle input.
Merged CRBD update needed to get same log Z in the CorePPL-TreePPL paper.
- Integrated new RootPPL syntax in CombineDS using NEXT
- ANADS - const nu
- conditionOnDetection and posterior sampling to CombineDS
Use rootppl
to compile an example
rootppl examples/CombineDS.cu
Use program
to run the executable
./program
For more details see the miking-dppl/rootppl
.