google bombs in the news! (including one or two I participated in, along with many other LJ-ers)

Indian scientists cure farts with radioactive beans. Think I’ll keep farting, thanks.

Scottish prisoners being given methadone to wean them back onto heroin before release. Yes, back onto heroin.

similar interest matching is back up. This feature allows you to enter any user name to receive a list of users with similar interests. This is only a paid members feature for now. Enjoy!

It didn’t do me a lot of good… I matched with a few people I like, and a few people that I don’t. Sad, but just because a person shares a lot of your interests, it doesn’t mean they write well, or share your opinions. In my “top 25”, there are two downright pleasant reads, two mehs, and and a lot of no thank yous.

I did find one or two that might be fun new reads, so I guess it worked…but all the groovies that I really like thus far I seem to already peek in on.

Interesting to see the index back though. I think it’ll be more useful for finding communities by single interests. Common interests don’t work as well when they’re too common… a like of “cookies” might attract a baker, a bulemic, and a web programmer. they won’t always have things to say that’ll interest one another. I will say that my first match has quite a bit to say that I enjoy reading heaps and heaps.

http://www.livejournal.com/interests.bml?mode=findsim

10 Commandments is on…. Moses was quite the badass, if I may use the term. He didn’t take any guff, nosiree. In my top three Chuck roles, with Planet of the Apes and Ben Hur. (The Omega Man and Soylent Green rocked, but in a *bad* way).

Suppose that the site had 2049 users. If I and 256 other users list “programming” as an interest, the information content in a match is log(2048/256) = 3 bits. If I and 16 other users list “perl” as an interest, the information content in a match to “perl” is log(2048/16) = 7 bits. These numbers are additive (in the ideal case where interests are chosen independently), so the total information content in the pair of matches is 10 bits (reflecting the fact that only 1 out of 210 users is expected to match both interests).

When ranking the combinations of matches, the units are unimportant, so logarithms of any base can be used. In the example, the information content can be computed most efficiently as 2*log(2048) – log(256) – log(16). So the algorithm is this:

Set logUserCnt = log(total # of other users)
For each of the primary user's interests:
Set offset = log(# of other users with that interest)
For each 'matchingUser' (another user with that interest):
If matching user isn't yet represented by an object
Construct a new MatchingUser object, initializing:
matchingUser.matchCnt to 0
matchingUser.cumOffset to 0.0
Increase matchingUser.matchCnt by 1
Increase matchingUser.cumOffset by offset
Rank matching users by similarity, computed as:
matchingUser.matchCnt*logUserCnt-matchingUser.cumOffset