cscheid

Sorry, but if that were the case, then just showing the list of pages and their locations would get 100% of them, with 0% probability of being "by chance".

Then, you say "that's a complicated rule: no one can memorize the list", and that's fine. But finding the simplest rule that has the best fit can be shown, via the Minimum Description Length principle, to be equivalent to models with deep connections to statistical learning.

The gist of the matter is: even if you have a single dataset, statistical considerations are very important to make sure that the statements you make are sensible. That does not appear to be the case in this Slate piece.

This kind of spatial scan statistics is really easy to misuse without careful multiple-comparison corrections. Would they have reported a 52% effect? a 51% effect? What about 1.4 inch bands? or 1.2 inch bands? That "remarkably thin" 0.3 percent will disappear pretty quickly.

Sorry that this is dismissive, but this piece is just a slightly more complicated case of http://xkcd.com/882/

Agreed!

What I'm trying to get at is that there's two ways you can think of academia: 1) roughly the set of people in traditional institutions, and 2) a basically unorganized set of people that happen to come together through working on related problems, in related ways.

2) is arguably the purer view of academia, and in that sense, it's as strong as ever.

I agree with you that there's deep, serious problems with 1), and they have to do with Pournelle's law, (in the US) with Bayh-Dole, and a litany of other hard, nasty problems of social life. But, even in that case: is that any worse than other cultures?

John Baez has had an interesting series of posts about this on G+, but IIRC the gist of it is that even though both Grigori Perelman and Yitang Zhang were somewhat outside the mainstream of academia, their work was not. Both of them built on well-known theories and methods to prove their groundbreaking theorems. Perelman's primary tool, for example, was Ricci flows, which had been used at least 25 years before his work. (I know even less about Zhang's work)

It's not cutting off. It's replacing many long-range effects with a single one weighted more (and done it carefully so that the floating-point error is literally smaller than you could tell).

There's even better methods that take O(n) time, like the fast multipole expansion ones.

For one, because if someone does find a hole that gives them access to Github data, they'll have all password attempts, which would include typos of the real password. Which is a terrible, terrible thing to store in a hard drive (see Adobe)

I recently scored a copy of Maxwell's Demon, a book containing a collection of (among others) Landauer and Bennett's papers. As I was skimming them, I couldn't help but think that if there is a correspondence between heat and information erasure, can that be used to move "heat" in a fundamentally different way?

The idea would be you do some computation at point A, the related information erasure at point B, and the heat associated with it get "transported" faster than it would be by moving pipes around. The hope is that sending the bits along something like an optical network would be faster than laminar fluid flow.

I'm sure HN's more knowledgeable folks will tell me exactly where I'm wrong, but it seemed like an intriguing possibility.

And one could use Teddy and the myriad followup papers to automatically get the 3D profiles:

http://www-ui.is.s.u-tokyo.ac.jp/~takeo/teddy/teddy.htm

The important thing in this demo is that his code reduces to arcball-like rotation in the globe projection, but works on arbitrary map projections. Notice how the interaction works for Eisenlohr's projection in the same way as the globe, but the behavior must be different, of course, since the projections are so unlike each other.

> > fantastic [design] work...

> (Someimes good enough ...

Can you see the problem here?

The number your comparison is missing, and one that happens to be crucial to both business and technical differences, is that you should be looking at tweet impressions, not tweets.

Seems like a rare case where using this technique gives you much more entropy from a good passphrase than from a good password (although I guess over the shoulder watching is similar)

And then there's this, about that paper: http://junkcharts.typepad.com/junk_charts/2010/05/8-red-flag...

Edit: And this http://junkcharts.typepad.com/numbersruleyourworld/2010/05/m...

We've been working on an iPython-notebook-like system for R, called RCloud: http://github.com/cscheid/rcloud. It's not nearly as full-featured as you want, and still very heavily in development. Still, it does R-to-Javascript RPC (and back), allows you to develop and share your R scripts directly in a browser, and provides a rudimentary way of doing interactive Javascript plotting (so you can use write code that uses d3 to visualize R data, for example).

Similarly, I once heard a story that http://en.wikipedia.org/wiki/David_(Michelangelo) has a bigger head than usual because people would be looking at it from a funny perspective, and foreshortening would make it look strange if the proportions were not changed.

So I looked it up, and it's actually on the seminal 1948 paper:

http://cm.bell-labs.com/cm/ms/what/shannonday/shannon1948.pd...

It turns out that I either hallucinated (incorrectly, perhaps) one side of the argument, or I read it from some other place, because the paper only talks about one side, the one about feasibility. The relevant paragraph follows:

"The redundancy of a language is related to the existence of crossword puzzles. If the redundancy is zero any sequence of letters is a reasonable text in the language and any two-dimensional array of letters forms a crossword puzzle. If the redundancy is too high the language imposes too many constraints for large crossword puzzles to be possible. A more detailed analysis shows that if we assume the constraints imposed by the language are of a rather chaotic and random nature, large crossword puzzles are just possible when the redundancy is 50%. If the redundancy is 33%, three-dimensional crossword puzzles should be possible, etc."

Anyway, to somewhat rescue a botched story, the ambiguity part reminds me of the mindblowingly awesome and famous puzzle by Jeremiah Ferrell on the eve of the 1996 presidential election, which clue was "Lead story in tomorrow's paper", with possible answers BOBDOLE ELECTED or CLINTON ELECTED, both satisfying all crossclues as well! Amazing:

http://cdn0.dailydot.com/uploaded/images/original/2013/2/12/...

Reminds me an argument about crossword puzzles I think comes from Shannon.

It goes like this: in a sufficiently "fault-tolerant" language (ie with low per-character information content, which sorta gives large Hamming distance), crosswords become impossible, because puzzles won't be satisfiable: words could never line up right. But in a sufficiently compact language (many bits per character in words, and Hamming distance thus tending towards zero), crosswords are impossible as well, but now because puzzles would be ambiguous: too many words that fit. Somehow, natural languages appear to fit somewhere in the middle.

The slickest part of that demo is the continuous level-of-detail rendering. That's extremely helpful when scaling up to large maps.

It's more than assignments looking normal. "await x" is better since it's an expression and not a statement. That is, "await x" will be a placeholder for a value (after the appropriate CPS, of course), while the assignment syntax forces you to use a temporary variable, even if you don't need it.

(for whatever's worth, I'm Brazilian)

Holy Baader-Meinhof, just today, in frustration, I wrote something like Haskell's sequence_ for ContT, in Javascript:

https://gist.github.com/cscheid/6241817

If you drop stage 1 in Earth's gravity well instead of Mars's.. yes it does :)

Edit: as mikeash points out in the reply, I'm plain wrong. You could use multi-stages to accomodate different gravities when launching Earth-Mars-Earth (2 stage E->M, reusable as standalone second-stage, so single stage M->E).

I'm quoting you:

> I'm convinced anyone can do this and the major thing stopping them is their belief that they can't do it.

It's not a matter of taking offense. It's a matter of your assumption. It took you little effort. And it might take many people little effort. But by saying "anyone" and "their belief that they can't do it" in the same sentence you disenfranchise people who, you know, might actually have tried this and failed.

This type of statement tends to irk some people (and I'm here assuming sillysaurus is part of this group) because the same structure is present in all kinds of victim-blaming statements ("oh, you don't work hard enough, that's why you're poor. You have to believe in prosperity").

Let me be clear, I'm not saying that you'd actually use the more horrible versions of the pattern. I'm simply hoping to explain sillysaurus's strong reaction to your original post.

"Wednesday, August 14, 2013 Last Update: 11:07 AM ET" 2 hours ago as of this post.

"Willpower" (http://www.amazon.com/Willpower-Rediscovering-Greatest-Human...) calls what you're describing "precommitment", and claims it does work consistently well.

Well, you don't know that. It could have been 0.16% instead of 1.6%. All we know is the math doesn't work.

(Replying here because your other comment is too deep)

I believe you have it right. The NSA document gets it wrong by one digit (page 6):

http://i2.cdn.turner.com/cnn/2013/images/08/09/2013_08_09_th...

(0.025/100) * (1.6/100) = 0.0004%, and not 0.00004% as claimed in the PDF.

I remember reading (~10-15 years ago?) that although the general public's impression was that propellers make planes unreliable, it was actually turbines that made airplane engines reliable (which means you really want turboprops rather than piston-propeller aircraft)

Have airplane piston engines become as reliable as turboprops?

This is very cool!

A very nice feature to have would be to run topic modeling (say, your run-of-the-mill LDA) on the comments, and then colormap the nodes as to preserve the distances in the comment topic vector. This way you'd be able to see threadjacking, etc.

The panopticlick paper, https://panopticlick.eff.org/browser-uniqueness.pdf, suggests a JavaScript library called PluginDetect, http://www.pinlady.net/PluginDetect/