I don’t think so, provided you limit yourself to Turing machines constructed of at most a universe’s worth of energy and have unlimited time for the computation. However, that said, invoking such a heavy-duty prior seems to me a bit of a cop-out; you might almost as well invoke “My prior is what Omega told me”. For choosing between Bayesian and frequentist approaches regarding actual problems in the real brain space available to humans, I prefer priors that can be computed in something like a human lifetime. To show that a Bayesian with a prior that would take multiple billion years to compute can do better than a frequentist is not really a very strong result.
I don’t think so, provided you limit yourself to Turing machines constructed of at most a universe’s worth of energy and have unlimited time for the computation. However, that said, invoking such a heavy-duty prior seems to me a bit of a cop-out; you might almost as well invoke “My prior is what Omega told me”. For choosing between Bayesian and frequentist approaches regarding actual problems in the real brain space available to humans, I prefer priors that can be computed in something like a human lifetime. To show that a Bayesian with a prior that would take multiple billion years to compute can do better than a frequentist is not really a very strong result.