I'm not totally convinced of the importance of universal priors. Any thoughts on this?
-
-
Replying to @ObjectOfObjects
Well, Garrabrant inductors are better than universal priors, and they're computable. So.
1 reply 0 retweets 1 like -
Replying to @ProofOfLogic
I don't know much about Garrabrant inductors. Can you use them for sequence prediction? And they're not universal?
1 reply 0 retweets 0 likes -
Replying to @ObjectOfObjects
You can. They eventually approximate a universal inductor in a technical sense specified in the paper (iirc).
1 reply 0 retweets 0 likes -
Replying to @ProofOfLogic @ObjectOfObjects
You can think of it as a universal approximation of a universal prior, in that it does at least as well as any poly time approximation would
1 reply 0 retweets 0 likes -
Replying to @ProofOfLogic @ObjectOfObjects
(It takes exp time to dominate all poly time algs, so it's not dominating the same class; but that's provably impossible for this task)
1 reply 0 retweets 0 likes -
Replying to @ProofOfLogic @ObjectOfObjects
But it doesn't converge to a universal prior exactly, it just converges to a prior which strictly dominates the universal semimeasure (iirc)
2 replies 0 retweets 0 likes
Which, I'd say, is even better. But the good properties at finite time are the real reason to favor it.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.