< Previous
Next >

: And unless I missed something, which I probably did, I'm bringing CRSMITP to a close for now with stuff from Xorph's Brendan Adkins. He sent me an Ambiguous Headline which he promptly put up on his own weblog so I can't use it, but earlier he responded to my discussion of least-relatedness, which I can and will use.

The least-relatedness Google feature would be GREAT, assuming it was in any way related to content (which I don't think it could help but be). You could search for, say, Cartoons by Gaspirtz, and then get whatever sites were least related to it, and you'd have a Best of the Web list right there!

He marks this in <easy-jab> tags, but since I am pedantic to the point of aggrevation I must point out that relatedness is orthogonal to quality. A good comic like Brendan's is more related to Gaspritz's work than is the boring web page about my cat (it's so boring, I don't even have a cat).

He also claims that you could find the worst path of a graph by having a greedy algorithm grab the longest edges instead of the shortest edges, but 1) that doesn't work when all the edges are the same length because they merely represent links between sites, which was my working assumption, and 2) it doesn't work anyway, for the same reason a greedy lowest-cost-path algorithm doesn't work.

A--(1)-->B--(1000)-->D
|                    ^
+--(2)-->C----(1)----+

A greedy algorithm chooses ACD, but ABD is much better. Brendan says "I suck at discrete", so a tip: it's usually easy to come up with counterexamples using extreme cases like the one above.


[Main] [Edit]

Unless otherwise noted, all content licensed by Leonard Richardson
under a Creative Commons License.