6533b854fe1ef96bd12af20f

RESEARCH PRODUCT

Componentwise adaptation for high dimensional MCMC

Eero SaksmanHeikki HaarioJohanna Tamminen

subject

Statistics and ProbabilityMathematical optimization010504 meteorology & atmospheric sciencesMonte Carlo methodMarkov processMarkov chain Monte Carlo01 natural sciencesStatistics::Computation010104 statistics & probabilityComputational Mathematicssymbols.namesakeMetropolis–Hastings algorithmTest caseChain (algebraic topology)Component (UML)symbolsStatistics::MethodologyErgodic theory0101 mathematicsStatistics Probability and Uncertainty0105 earth and related environmental sciencesMathematics

description

We introduce a new adaptive MCMC algorithm, based on the traditional single component Metropolis-Hastings algorithm and on our earlier adaptive Metropolis algorithm (AM). In the new algorithm the adaption is performed component by component. The chain is no more Markovian, but it remains ergodic. The algorithm is demonstrated to work well in varying test cases up to 1000 dimensions.

yearjournalcountryeditionlanguage
2005-06-01Computational Statistics
https://doi.org/10.1007/bf02789703