Multiphase Monte Carlo Sampling #147
Conversation
R-proj/src/copula.cpp
Outdated
| @@ -7,6 +7,7 @@ | |||
|
|
|||
| //Contributed and/or modified by Apostolos Chalkis, as part of Google Summer of Code 2018 program. | |||
|
|
|||
|
|
|||
There was a problem hiding this comment.
many files have this empty line insertion, could you undo these changes?
There was a problem hiding this comment.
done in root/R-proj folder
|
|
||
| NT variance = (normalized_samples.cwiseProduct(normalized_samples)).sum(); | ||
| variance *= (1.0 / N); | ||
| variance += NT(1e-16)*(samples_mean*samples_mean); |
There was a problem hiding this comment.
what is the purpose of this computation? why 1e-16? Is this some magic number that simply works?
There was a problem hiding this comment.
Yes, it was added to avoid numerical overflow in case where (normalized_samples.cwiseProduct(normalized_samples)).sum() is zero.
Now I use const NT eps = 1e-16;
| @@ -173,15 +248,15 @@ struct AcceleratedBilliardWalk | |||
| T -= _lambda_prev; | |||
| P.compute_reflection(_v, _p, _update_parameters); | |||
|
|
|||
| while (it <= 100*n) | |||
| while (it <= 1000*n) | |||
There was a problem hiding this comment.
what is the reason behind this change? stability?
There was a problem hiding this comment.
Better performance in the initialization of the parameters when we have very skinny polytopes.
Now I use unsigned int _rho = 1000*n;
include/sampling/mmcs.hpp
Outdated
| > | ||
| void perform_mmcs_step(Polytope &P, | ||
| RandomNumberGenerator &rng, | ||
| const unsigned int &walk_length, |
There was a problem hiding this comment.
should this be unsigned int const& as below ?
include/sampling/mmcs.hpp
Outdated
| unsigned int &total_samples, | ||
| unsigned int const& num_rounding_steps, | ||
| MT &TotalRandPoints, | ||
| bool &complete, |
There was a problem hiding this comment.
I think is is clearer if this functions returns this bool and below you simply return true or false instead of setting this variable and return void.
include/sampling/parallel_mmcs.hpp
Outdated
| unsigned int &total_samples, | ||
| unsigned int const& num_rounding_steps, | ||
| MT &TotalRandPoints, | ||
| bool &complete, |
This PR implements a Multiphase Monte Carlo Sampling (MMCS) algorithm that unifies sampling with distribution isotropization. The PR contains:
-- A C++ implementation of a single step of MMCS
-- A C++ class to update the Effective Sample Size for each new generated chain
-- A C++ implementation for the estimator of autocovariance in "Charles J. Geyer, Practical Markov Chain Monte Carlo, Statistical Science 1992."
-- New burn-in routines for Billiard Walk.
-- Two C++ examples to illustrate the new method.
This PR depends on #142 and #144