Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
* Move expr_builder inside expression * In progress of modifying var vec mat * Separating param/data with var/vec * Draft of data and param with corresponding traits * Refactored everything except mcmc including tests * Add integration test for subscripting vec-like params * Rename to integration test and combine all expression tests into one executable * Add support for metropolis hastings * Restore test cases to original sample size * Add support for NUTS * Fix small syntax in examples * Fix uninitialized value issue in sampler_tools test * Try generating and saving the value * Try initializing member arrays in ctor * Using c++20 style implementation of identity * Activate model! * Add test cases for min and max edge cases * Resolve some TODOs * Clean up traits related files * Fix syntax for storage * Fix pdf to have same API as log_pdf * Fix header includes * Add dot product, cached AD vars, stronger concepts - More type safety has been added in traits - Model ad_log_pdfs and variable expression to_ad member functions allow users to pass in cache in case these expressions need them. - Dot product finally implemented (performance is still great) - NUTS creates long cache vector (the extra memory for adj and values may not be necessary) - more unittests with dot product * Optimize NUTS to allocate less memory for cache variables * Reorganized examples, write up design doc, and clean up type safety in binop * More complicated regression example * Add more comments and complete type check on all classes. Fix really really important bug in dot * Move activate and activate_cache outside of mcmc * Modify test to use cache * Apply caching for normal and uniform * Modify mcmc algorithms with different activate API * Add support for bernoulli ad expression * Remove gcc no warning flag and static cast void unused lambdas * Add python script to compute true answers for nuts unittests and modified to have universal tolerance lvl * Add support for normal diag covariance * Change seed and reset tol to 0.05 * Add more comments and higher tol * Increase tol even more * Change seed to 1 and tol to 0.7 * Change seed to 0 and tol to 0.7 * Remove universal tol lvl * Increase tol for std normal and stddev vector case * Add autocorrelation * Add improved autocorrelation statistic and finish ESS * Add ESS information in regression ex * Add overload of ess for 1 chain case * Fix expression builder for operator, and operator|= and dot to convert params to viewers * Modify NUTS to output timed results and finish README!! * Add disclaimer * Change some numbers to hit an if statement
- Loading branch information