Metropolis-Hastings Buddhabrot

[PurpleBlu3s]

Hi, I've come back to trying to generate some buddhabrots, and I really want to get some different levels of zoom. I gather that the best way to do this is by using the metropolis-hastings method, but I think I'm not quite understanding it properly as my attempts so far aren't even close to working.

Could anyone point me in the right direction / explain in simple terms how to go about coding it.

Thanks.

[lycium]

Alex Boswell has written about it here: http://www.steckles.com/buddha/

I've found that a much simpler implementation works well, too.

[PurpleBlu3s]

In what sense simpler?

[lycium]

The mutation function can be made a bit simpler, and it seems to still work well. However, M-H sampling should only be used for reasonably deep zooms, as it can be easily outperformed when rendering a large fraction of the set (M-H sampling is a form of rejection sampling, which always increases variance).

I couldn't find any zooms in the set that looked particularly interesting when I tried it back when.