Hacker News: nphrk

New comment by nphrk in "The Central Limit Theorem Visualized with D3"

nphrk — Thu, 30 May 2013 20:55:06 +0000

It is the distribution of the average of n binomial distributions (taking on values x±1)†, which according to the CLT converges to a Gaussian as n→∞.

[†] x is where it is centered.

New comment by nphrk in "Instagram says it now has the right to sell your photos"

nphrk — Tue, 18 Dec 2012 15:25:55 +0000

The copyright issue can be a bit tricky. Assuming that Instagram goes on and sells users' pictures - who's liable if somebody uploads a picture which he/she doesn't own, and then Instagram goes on and sells that to a third party?

New comment by nphrk in "Prismatic gets $15 million in Series A Funding"

nphrk — Wed, 05 Dec 2012 19:05:58 +0000

Great! I love the service, I miss only an app for my Android tablet. Keep up the good work.

New comment by nphrk in "Understanding The Fourier Transform"

nphrk — Sun, 02 Dec 2012 21:21:29 +0000

The article uses the magnitude of the coefficients, which is computed using both the real and the imaginary part.

New comment by nphrk in "Understanding The Fourier Transform"

nphrk — Sun, 02 Dec 2012 21:19:29 +0000

This is a nice way to see how the DFT is computed, however I find the view of the FT as a change of basis as even more important - generalizes easily to other bases and and one can understand easily wavelets and their advantages. Basically, the sinusoids form a basis of the vector space of functions (every 'non-pathological' function can be written as a possibly infinite sum of them) and the numbers computed by the FT are coefficients for the respective basis vectors - the magnitude of these coefficients is interpreted as the strength of the corresponding wave in the original signal.

Another way to see the FT is as the basis where the convolution operators are diagonal - this is used in image processing, where computing the FFT of a filter + entry-wise multiplication can be much faster than running the convolution at each pixel of the input image.

New comment by nphrk in "Ask HN: what was the best life/programming choice you ever made?"

nphrk — Sun, 02 Dec 2012 14:52:09 +0000

Could you recommend some introductory books? I've always felt bad not having clue about art.

New comment by nphrk in "Blaze: Next Generation NumPy"

nphrk — Sat, 27 Oct 2012 21:11:12 +0000

There's http://www.videolectures.net, but I think that they only host their own material.

New comment by nphrk in "Red Bull Stratos Skydive Rescheduled for today"

nphrk — Sun, 14 Oct 2012 18:19:15 +0000

Just landed!

New comment by nphrk in "Infographic: US campaign finances revealed"

nphrk — Thu, 11 Oct 2012 15:45:50 +0000

Very interesting, I have only one nitpick (maybe for the paranoid only) : why not show the cumulative distribution but pick 200$ as a threshold?

New comment by nphrk in "Neural Networks for Machine Learning"

nphrk — Wed, 03 Oct 2012 12:27:53 +0000

Well not quite. While SVMs gained a lot of popularity for having nice properties e.g.

1) a convex problem which means a unique solution and a lot of already existing technology can be used

2) the "kernel trick" which enables us to learn in complicated spaces without computing the transformations

3) can be trained online, which makes them great for huge datasets (here the point 2) might not apply - but there exist ways - if someone's interested I can point out some papers)

There is an ongoing craze about deep belief networks developed by Hinton et al. (who is teaching this course) who came up with an algorithm that can train them reasonably well (there exist local optima and such, so it's far from ideal). Some of the reasons they're popular

1) they seem to be winning algorithm for many competitions / datasets, ranging from classification in computer vision to speech recognition and if I'm not mistaken even parsing. They are for example used in the newer Androids.

2) DBNs can be used in an unsupervised mode to _automatically_ learn different representations (features) of the data, which can be then used in subsequent stages of the classification pipeline. This makes them very interesting because while labelled data might be hard to get by, we have a lot of unlabelled datasets thanks to the Internet. As what they can do - see the work by Andrew Ng when they automatically learned a cat detector.

3) DBS are "similar" to biological neural networks, so one might think they have the necessary richness for many interesting AI applications.

New comment by nphrk in "Neural Networks for Machine Learning"

nphrk — Wed, 03 Oct 2012 04:25:27 +0000

Well not quite. While SVMs gained a lot of popularity for having nice properties e.g.

  1) a convex problem which means a unique solution and a lot of already existing technology can be used
  2) the "kernel trick" which enables us to learn in complicated spaces without computing the transformations
  3) can be trained online, which makes them great for huge datasets (here the point 2) might not apply - but there exist ways - if someone's interested I can point out some papers)

There is an ongoing craze about deep belief networks developed by Hinton (who is teaching this course) who came up with an algorithm that can train them (there exist local optima and such, so it's far from ideal). Some of the reasons they're popular

  1) They seem to be winning algorithm for many competitions / datasets, ranging from classification in computer vision to speech recognition and if I'm not mistaken even parsing. They are for example used in the newer Androids.
  2) They can be used in an unsupervised mode to _automatically_ learn different representations (features) of the data, which can be then used in subsequent stages of the classification pipeline. This makes them very interesting because while labelled data might be hard to get by, we have a lot of unlaballed datasets thanks to the Internet. As what they can do - see the work by Andrew Ng when they automatically learned a cat detector.
 3) They're "similar" to biological neural networks, so one might think they have the necessary richness for many interesting AI applications.

New comment by nphrk in "Rod Johnson joins Typesafe's Board"

nphrk — Mon, 01 Oct 2012 13:33:48 +0000

There's a Coursera course by Martin Odersky going on at the moment (https://class.coursera.org/progfun-2012-001/class/index).

New comment by nphrk in "On{X}: The Coolest Thing to Happen to Android. Courtesy of… Microsoft Israel?"

nphrk — Tue, 05 Jun 2012 19:17:14 +0000

I guess you haven't seen the Old Spice commercial [ http://www.youtube.com/watch?v=owGykVbfgUE ].

New comment by nphrk in "20 lines of code that beat A/B testing every time"

nphrk — Tue, 29 May 2012 22:11:29 +0000

There are better approaches for tackling this problem (with 0-regret asymptotically). You can take a look at the UCB (Upper Confidence Bound) algorithm, and you can do even more if you assume some continuity, e.g. what is commonly done is to assume that the whole distribution is from a Gaussian Processes. Many interesting ideas in the literature indeed :)

New comment by nphrk in "Show HN: Facebook stockvalue (prettified)"

nphrk — Tue, 22 May 2012 12:23:48 +0000

Looks nice I must admit :) Are you fitting a quadratic curve? When you have multiple points what do you plan to use - a spline?

New comment by nphrk in "Google Summer of Code 2012 Stats - Part 2"

nphrk — Sun, 20 May 2012 20:26:53 +0000

No US universities, I guess they can easily get way better paying internships.

New comment by nphrk in "Kindergarten Teacher Earns $700,000 by Selling Lesson Plans Online"

nphrk — Fri, 18 May 2012 08:05:46 +0000

Unfortunately, even if illegal, there is a way around it. So what some of the teachers where I studied did (one of the eastern European countries) was to send her/his students to visit tutorial classes of the other teacher (for which you pay) and vice versa.

New comment by nphrk in "Lenovo refreshes its ThinkPad T, W, L and X"

nphrk — Tue, 15 May 2012 19:01:54 +0000

Oh my, why the keyboard? Now I have one less reason to go with a Thinkpad over a MacBook :(.

2012: The Year of the Semantic Web

nphrk — Wed, 01 Feb 2012 14:22:12 +0000

Article URL: http://www.huffingtonpost.com/steve-hamby/semantic-web-technology_b_1228883.html?ref=tw

Comments URL: https://news.ycombinator.com/item?id=3538003

Points: 4

# Comments: 0

New comment by nphrk in "Stanford profs from DB & Machine Learning class are founding a company Coursera"

nphrk — Sun, 29 Jan 2012 16:50:29 +0000

You have some complaints about the ML class?