The essential point is that we're considering classification problems, for which the output is intended to be 0 or 1. I address the more general case of regression problems (where y may take any value) in a later exercise.
Hope that helps!
8,027 karma · joined July 24, 2007
The essential point is that we're considering classification problems, for which the output is intended to be 0 or 1. I address the more general case of regression problems (where y may take any value) in a later exercise.
Hope that helps!
http://johnen.shinshu-u.ac.jp/~mano/profile_e.html
The letter is from an excellent collection of Feynman letters, "Perfectly Reasonable Deviations from the Beaten Track". There's a thoughtful review of the entire collection (by Freeman Dyson) here:
http://www.nybooks.com/articles/archives/2005/oct/20/wise-ma...
The HN search results suggest that this is the second time the book has made it to the HN front page.
However, you may have picked up a different impression because the second chapter -- a largely self-contained introduction to backpropagation -- also made it once onto the front page. Given that I'm rolling out chapters as I finish them, this seems to me like multiple posts from the same blog appearing on the front page. Certainly, the individual chapters are (I hope) far more substantive than most blog posts.
(For the record, I had no idea this would be posted.)
The quantum search algorithm is very interesting. However, most people in the field would, I think, agree that a non-oracle separation between quantum and classical would be vastly more interesting. Oracle results feel a little too close to cheating.
There is, incidentally, a nice paper showing that for computing total Boolean functions the quantum oracle model can be no more than polynomially faster than a classical model. This means there can be no exponential speedup for such functions, which is a real pity.
http://rjlipton.wordpress.com/2013/04/27/sex-lies-and-quantu...
Yes, 't Hooft and others, including Leonid Levin and Oded Goldreich, believe that quantum computers won't function because of quantum mechanics breaking down. See, e.g.:
A year later detectives arrived at our house. They asked to speak to my mother, and told her that the house next door had actually been stolen, and the real owners were now looking for leads, hoping to find their stolen house.
As an example, "Inventing on Principle" has inspired many people and companies to think more deeply about development environments, and to try implementing more speculative ideas. A skeptic can point out that much of that consequent work is of mixed quality, but that's just Sturgeon's Law. What matters is that more people will think better thoughts about more important problems. And that's certainly occurring as a result of Victor's talks and essays.
This limit doesn't immediately impact computations that are carried out entirely on-chip. I expect that's the great majority of computations on today's CPUs, which typically have large caches. So I'm not sure the GP's explanation is correct, although it's certainly interesting.
On not using a conventional publisher: my first two books were with mainstream publishers (Cambridge University Press and Princeton University Press). For my third book I decided to self-publish online only, because that makes it possible to integrate animations, interactive Javascript, and so on. Four publishers have asked about publishing paper editions of the book, but it's hard to imagine going back to the printed page as the primary target medium. Having the web as the primary target is so much more powerful.
+ Yes, the code presented "works" for multiple hidden layers. But it converges too slowly to be useful, except for very small networks. Later chapters introduce new ideas which help backprop converge faster, and which start to make multiple hidden layers quite feasible.
+ On the broader machine learning front, I think scikit learn is a pretty nice general library: it's approachable, easy to use, and has reasonable docs. It'll be interesting to see how it develops.
+ For neural nets in particular, elsewhere in this thread Dave Sullivan has done a nice job distilling out a list of good libraries. I think it'll be a while before there's a real out-of-the-box solution for neural nets, though, since setting parameters for neural nets is something of an art.
(I really enjoyed reading the remainder of your comment.)
(Dropping some credentials doesn't seem inappropriate here: I worked professionally as a researcher on quantum computing and quantum information for 13 years.)
"It is not the critic who counts; not the man who points out how the strong man stumbles, or where the doer of deeds could have done them better. The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood; who strives valiantly; who errs, who comes short again and again, because there is no effort without error and shortcoming; but who does actually strive to do the deeds; who knows great enthusiasms, the great devotions; who spends himself in a worthy cause; who at the best knows in the end the triumph of high achievement, and who at the worst, if he fails, at least fails while daring greatly, so that his place shall never be with those cold and timid souls who neither know victory nor defeat."
The critics often seem to be saying, implicitly, "_I_ could do it better". It's a claim of superiority. I guess it's beguiling, because criticism of this form is so much easier than genuine creation. I think many of the critics don't consciously realize this, since they've never created something of great worth, and have no idea how difficult it is.
I've known very accomplished people who could be haters. But usually accomplished people seem to be appreciative of creative projects, and understanding of flaws (while wanting them eliminated). I think it's because they understand just how difficult it is to do good creative work. They know that someone can bring immense gifts and effort to bear on a project, and still have it go awry.
The companies may not benefit, but the developers certainly will! Not just the ones being moved around, either --- it'll raise market prices, and so _every_ developer will benefit, even those who don't move.