Re: Final(?) word on the No Free Lunch theorem


Bobby D. Bryant schreef:

> I've been waiting a long time to post this, but it seems that our
> resident evolution deniers have lost interest in the NFLT as an
> anti-evolution proof, so I'm posting it now without prompting rather
> than waiting for a context that may never arise.
>
> (Of course Pagano reflexively invokes the NFLT as if it were a charm
> of protection against evolution, but his assertions don't actually
> appeal to the content of the theorem, so there's no point in
> explaining it to him. He could equally well use *any* theorem, so
> long as it has a name that can be plugged into the sentence -
> "The _____ theorem disproves evolution.")
>
>
> But we were talking about the NFLT theorem. I think I have finally
> fully grokked it.
>
> The theorem seems counter-intuitive when you read it, e.g. the bit
> about a hill-climbing and hill-descending offering the same expected
> value for their performance.
>
> The problem is that the NFLT is cast as a theorem about search
> algorithms, and when we think about it we bring in all our baggage
> about how search algorithms should work. But your intuitive qualms
> about its counter-intuitive claims go away when you strip away the
> notion that it's about search algorithms and consider what it _really_
> says:
>
> You can't guess a random number on the basis of the previous
> random numbers you've seen.
>
> IOW, the theorem is merely a formal proof of a special case of what
> Hume told us two and a half centuries ago.
>
> And the thing is, we don't try to apply search algorithms to the
> problem of guessing random numbers, unless of course we don't
> recognize the problem for what it is.
>
> I find the theorem all but irrelevant to the field of search
> algorithms, to say nothing of its misapplication as a disproof of
> biological evolution.

There's an pretty good rundown on this theorem at
http://www.no-free-lunch.org/

See also:
http://www.talkreason.org/articles/orr.cfm (reccomended reading!)

<quote>
In real-life situations, it is the performance on a specific landscape
that counts and this is where evolutionary algorithms routinely
outperform random searches and do so very efficiently, both when the
processes are targeted (as in Dawkins's algorithm -see [8]) and when
they are non-targeted (as Darwinian evolution is).
</quote>

The basic objection to any application in ToE of course being that
evolution isn't algorithmic, but heuristic and seeks no specific
target. Small, but essential differences.

Here's more:

http://www.no-free-lunch.org/Shal02.pdf

<quote>
Has Dembski succeeded in making ID intellectually respectable? No. Let
me not pull any
punches: Dembski's No Free Lunch is a poorly written piece of
propaganda and pseudomathematics.
</quote>

.

Relevant Pages
Loading