You also make brief mention that the HNFLT proof assumes a partition of the search space and thus cannot handle the overlapping targets. This is problematic because any target on the original search space will become an overlapping target on the multi-query search space. I think you are correct on this point. Thanks for bringing it to our attention. We've added a description of the problem as well as an alternate proof for the uniform case on overlapping targets to the pdf: http://evoinfo.org/papers/2010_TheSearchForASearch.pdf.Here are some thoughts:
- Here is an excerpt from Tom English's blog Bounded Science:
A "Darwin or Design?" podcast interview (read on while it loads in a new tab, and then skip to 7:52) with Marks includes this cute little exchange:Question: Are you getting any kind of response from the other side? Are they saying this is kind of interesting, or are they kind of putting stoppers in their ears? What's going on?
Answer: It's more of the stoppers in the ears thus far. We have a few responses on blogs, which are unpleasant,* and typically personal attacks, so those are to be ignored. We're waiting for, actually, something substantive in response.
I made this point on overlapping targets in emails, blogs, wikis over the last two years. Which begs the question: Who is putting stoppers in their ears? What are the odds that I'll hear of the authors of the paper themselves? I now got my second mentioning in this paper, can I expect a bottle of whisky, and a handsigned copy of a book? BTW: I prefer the excellent "Handbook of Fourier Analysis and Its Applications" over "The Design Inference: Eliminating Chance through Small Probabilities" - By pure coincidence the home-page of Evolutionary Informatics has changed. At least until May 19, 2012 there was a link from it to the sub-page errata. Sadly this link is missing now - and the page on errata doesn't seem to be up-to-date.
- I didn't look into the erratum in detail, but I suspect that the theorem Given a uniform distribution over targets of cardinality k, and baseline uniform distribution, the average active information will be non-positive should be phrased more trivially as Given a uniform distribution over targets of cardinality k, and baseline uniform distribution, the average active information will be zero
1. DiEb sent me copies of email notes, and I believe that Dembski and Marks must have understood well in advance of publication that the "Horizontal No Free Lunch Theorem" was terribly flawed.
ReplyDelete2. The erratum is itself erroneous in its arguments, though perhaps correct in its claims. I suspect that Ewert wrote it, given the confusion of probability measures with algorithms. There are prominent statements in the paper that contradict the main "no free lunch" theorem, and the erratum does nothing to identify and correct them. No one at the EvoInfo Lab is owning up to how terribly confused they are about NFL.
3. This echoes my experience with Marks, more than a year before his and Dembski's first publication. At that time, I was affiliated with his Evolutionary Informatics Lab, in protest of what I considered to be Baylor's infringement of his academic freedom. I thought that he was operating in good faith, and that he would respond appropriately to feedback from someone with more experience than he in NFL and evolutionary computation. I pointed out that his and Dembski's first article characterized Dawkins' Weasel program incorrectly, and showed clearly that it implements an evolutionary algorithm that has been analyzed (as a Markov chain) many times. He did not respond. He and Dembski analyzed (as Markov chains), in the same article, very similar algorithms, with alphabet {0,1} instead of {A, B, ..., Z, *}. The analyses had been in the literature I pointed Marks to for at least ten years. Furthermore, Marks had served, along with me, as a technical advisor for David Fogel's Evolutionary Computation: The Fossil Record, and had seen one of the algorithms in an early paper in evolutionary computation.
4. What these incidents have in common is that the papers were in review or had already been accepted for publication. Dembski and Marks do not care so much about getting things right as getting things published.
5. It's become clear that Marks was incensed by the outcome of the Dover trial. My guess is that he joined up with Dembski in order to prepare "scientific" evidence for the next judicial test of public-school instruction in ID creationism. IDC had a woefully short list of peer-reviewed publications in the Dover trial, so volume matters. Also, D&M are obviously planting loaded terminology and plain-language exposition in their papers that will, when quoted in expert testimony, give a cooperative judge good verbiage for his or her decision.