search

binarybana / SAMC.jl

Stochastic Approximation Markov Chain Monte Carlo in Julia.
Other
1 stars 0 forks source link

A proper REQUIRE file is needed #1

Open binarybana opened 9 years ago

binarybana commented 9 years ago

Also wonder if I should push to METADATA.jl?

goedman commented 9 years ago

@binarybana Yes, this gets me a lot further! Thank you!

Will study it a bit more.

binarybana commented 9 years ago

SAMC (the technique) definitely has a lot of tuning knobs that make it somewhat hard to use, but the samc_utils.jl file has a plotting script to try and help with some of those. To make sure the sampling is uniform across the energy space etc...

The most important parameters are stepscale (which controls the internal parameter evolution delta). And for population SAMC, beta is similarly important.

Feel free to ask questions here (or file another Tuning bug).

On Wed Jan 07 2015 at 11:54:30 AM Rob J Goedman notifications@github.com wrote:

@binarybana https://github.com/binarybana Yes, this gets me a lot further! Thank you!

Will study it a bit more.

— Reply to this email directly or view it on GitHub https://github.com/binarybana/SAMC.jl/issues/1#issuecomment-69061796.