### Max G'Sell - Adaptive sequential model selection for the graphical lasso

88
views
0
7 months ago by
Community: WHOA-PSI 2017

Would a regenerative MCMC approach help you here? In case you don't actually have independence of the $S^{*}$ samples, you could just generate a long chain, using a regenerative method to break up the chain into i.i.d. `tours'. You can do a block version of this to sample in multiple dimensions.