5 ms·
We assert that a randomized algorithm is unnecessary, and the problem can be solved deterministically (though of course random numbers will be used for sa
by Udo 13y ago
We assert that a randomized algorithm is unnecessary, and the
problem can be solved deterministically (though of course random
numbers will be used for sampling).
At a certain point this distinction becomes just philosophical wankery. Make no mistake, the proposed deterministic algorithm is a random one in disguise. And that's fine. Over large sample sizes, this randomness becomes sufficiently deterministic. Which is what the original proposal suggested, and it's also why sometimes a random variable can be elegant, too.
- larksimian 13y agoThe article's philosophical wankery was way too pragmatic. Random vs deterministic is a spectrum, anyway. Given enough information about the state of the machine doing the random picking/number generation all computer algorithms are deterministic anyway. Heck, given enough information about the state of the universe everything is theoretically predictable, some predictions are just really hard. Hopefully this achieves the needed philosophical wankery quota.
- Udo 13y agoThat's sort of what I was talking about. Sorry if that wasn't clear. Anyway, I'm with you on that. :)
- kens 13y agoThe article describes an algorithm that randomly shows articles based on a computed distribution, and calls that "deterministic". That doesn't sound deterministic to me - is the article misusing the term, or am I misunderstanding "deterministic"? (genuine question)
- yummyfajitas 13y agoThe algorithm deterministically computes how many people should be shown article A and B and randomly assigns people to those different groups. I suppose any algorithm involving sampling will be random in this manner, but the point is that randomness is not being introduced to work around flaws in the decision procedure.