Hacker News Reader

dododoon LulzSec hacks into Bethesda Softworks accessing 200k Brink user accounts

there's no evidence they rooted senate.gov. just local user access. maybe the coldfusion user?

dododoon Lodsys explains in-app patent warning letters sent to iOS developers

here's the actual explanation from lodsys: http://www.lodsys.com/blog.html

they mention apple, google, microsoft have some sort of license from them, i wonder if the relatively low cost of a license is the reason: http://www.lodsys.com/1/post/2011/05/q-what-are-you-charging...

i.e., cheaper to just pay them off than fight it.

dododoon Last.fm launches crowdsourcing experiment for tempo (BPM) detection

given that there is no stated end date to this experiment, when exactly will you post the data? ("once it's all in" doesn't say anything).

will you post all of the data (in its original, unadulterated form) collected for free, unrestricted download by anyone for any purpose, including competing with last.fm?

intending to share, and actually making a public promise to do so, are quite different things.

dododoon Neural Networks making a come-back?

backpropagation is just the chain rule of differentiation. each layer of an ANN is logistic regression, which statisticians have been and continue to be interested in.

i'm not sure your caricature of the netflix winning solution is correct: i believe it was a blending of around 25 different models (including, i think, the RBM someone pointed about above) each in themselves quite varied from one another. this is typically how these challenges are won.

dododoon Neural Networks making a come-back?

in machine learning, the 'big' thing about neural networks at the moment is deep learning via some sort of unsupervised pre-training. there's be a lot of work published on this recently.

http://deeplearningworkshopnips2010.wordpress.com/

dododoon LIBSVM - A Library for Support Vector Machines

you might also want to have a look at shogun:

http://www.shogun-toolbox.org/

it provides a nice wrapper around libsvm, liblinear, and a whole bunch of other classification libraries. plus it provides things like HDF5 support, octave, matlab, python and R bindings, more esoteric kernels (e.g., on strings) as well as one-class and multi-class SVMs.

dododoon SSH Key Gen/Install Script. Automates Mass Distribution of SSH Keys

no, you are using a unix-like system. if the user does not have root privileges, they cannot perform your ridiculous suggestion.

the correct way to elicit a password from the user for ssh is to use ssh-askpass. a safe way to pass this to processes such that it does not appear in the process table is to pipe it.

dododoon Overlooked book may decipher Quipu

  Despite the scientific impact of their discovery, both León
  and Vega agreed to end their research and allow other
  academics to take over. After five years working on the 
  project, they are ready to let it go and allow other 
  experts continue the research.

this seems bizarre to me. it cannot be a matter of funding. they've spent 5 years gathering a great deal of expertise, and now decide to quit just when momentum has built up? they could instead seek collaborations on future work, if they want other people to work on this direction. otherwise there is likely a lot of catching up to do that is perhaps intimidating.

dododoon Haskell’s Niche: Hard Problems

that's how you write haskell the first time. that's fine. easy, nice, simple. you've written a lot of code; it compiles, it works, no problem.

then you realise that you want to change a data structure or a function needs some information that you want to weave through it or you used the wrong abstraction. then it's a pain in the ass.

dododoon Haskell’s Niche: Hard Problems

i'd heavily used haskell for 5 years. i was taught by some of the people who developed the original haskell specification, and the people who worked on haskell prime. i aced their haskell course (~95%. first in the year, iirc).

i've used haskell in work i was paid for, i've used haskell for real-time low-level video processing, as well as high level machine learning. i've contributed to a few open source haskell projects. the entire code for one of my research projects was written in haskell, coming to 2.5k lines of code.

you'll often find people who advocate haskell have the attitude that you "just don't get it" or "don't have enough experience" (an attitude you might find purveys some of the responses to my comment... i certainly do). i do like haskell, it's beautiful in many ways, but for large, hard problems where it's tricky to know how to develop code, i found it inappropriate. where you know what you want to write (i.e., have a specification that is solid or are willing to spend a while properly engineering things), haskell is great.

dododoon Haskell’s Niche: Hard Problems

   Haskell shines at making frequent fundamental changes.

my experience of haskell is completely the opposite of this. haskell is terrible when you want to make big changes---it's often quicker to just rewrite completely from scratch. fixing all the little inconsistencies is a real pain. want to add an extra parameter to a type class? want to remove an attribute from a data type? sure, if you picked the right abstraction from the start, this is a non-issue, but that's not how it is with a hard problem.

the type system gets in your way; this is why i now use python (colleagues have noted the same problems when using c++; we prototype in python, scale up in c for the slow bits). often when i'm fleshing out ideas or trying a new approach, i don't need everything to be consistent---just the bit i'm working on, even then i don't care too much. if things don't quite work, i'd like a quick way to find an inconsistency (often i write a consistency checker function for each class that i use when debugging, that checks types and structures).

i find haskell fundmentally useless for tackling hard (research) problems for this reason alone.

dododoon Show HN: What Do You Think Of My Resume?

i cannot read it. apparently you need a facebook account to see it.

dododoon Why text messages are limited to 160 characters

the mean isn't a robust statistic. i don't think your inference is correct in this case.

a single user can theoretically throw the mean off from the central tendency completely. i would imagine there are a few users that send a lot more text messages and calling more than most people, biasing the result. it would be more useful to talk about the median in this case, i think.

http://en.wikipedia.org/wiki/Robust_statistics#Examples_of_r...

dododoon Rovio explains why Angry Birds update needs SMS permission

i wish i could refuse to give android apps certain permissions yet still install them. i don't want angry birds to ever be able to charge me money and certainly not read all my SMS messages. i simply won't upgrade.

for other apps, i just have to not install them; it doesn't seem like i can give feedback to app developers about this.

dododoon Show HN: EDW, quantitative analytics, machine learning.

if you use some kind of regularization (e.g., if you think the parameters are sparse) it's possible to fit to such a large space without overfitting. this is common in machine learning and statistics (e.g., "N less than p" problems, where N is the number of data and p is the dimensionality of parameterization, common in genomics).

also provided the test data (i.e., the data it's bidding on) is not used in training, this should be a fair(ish) test; overfitting on the training data should lead to poor test performance.

however it's not clear to me that the data aren't being used twice, and what machine learning is actually going on... so in the best of worlds it could be ok, but given the scant details it could be all bunkum...

dododoon Simple Haskell webapp: Generate random tweets using Markov chains

are you using smoothing for large n? kneser-ney smoothing seems to give the best results.

http://nlp.stanford.edu/~wcmac/papers/20050421-smoothing-tut...

dododoon Can Google Get Its Mojo Back?

from the article...

   the data is compelling

...

   Even their money fount, AdWords, is problematic. An
   illustrative anecdote: I ...did this and that...

...

   While I can’t quantify this, I’m confident that most engineers will agree:

the article claims to argue from data, but presents no actual data: simply anecdotes. c.f., http://www.stat.columbia.edu/~cook/movabletype/archives/2011...

dododoon The Learning Behind Gmail Priority Inbox [pdf]

right, a key point of this method is that they are doing transfer learning from the general gmail population to each individual user. the reason for this is that the data from each user is small and sparse, and has relatively little statistical power.

you simply cannot do this on a desktop system unless you pull in data from other users. you might be able to do that but: bandwidth, privacy concerns, and you might find you have to do just as much work to get the same performance.

dododoon The Learning Behind Gmail Priority Inbox [pdf]

this work was presented as a mini-talk at the NIPS conference workshop on "Learning Cores, Clusters and Clouds":

http://lccc.eecs.berkeley.edu/papers.html#mini_talks_i

dododoon Hello World in LLVM

  $ cat >hello.s <<EOF
  .text
  .globl main
  main:
  	leaq msg, %rdi
  	callq puts
  	ret

  .data
  msg:	.asciz "hello world\n"
  
  EOF
  $ as -o hello.o hello.s; cc -o hello hello.o; ./hello
  hello world

dododoon Every "best paper" from Computer Science conferences since 1996

you can find the previous icml best paper awards by looking at the previous conference websites. for example:

http://www-ai.ijs.si/SasoDzeroski/ICML99/award.html via http://www.machinelearning.org/icml.html

in machine learning, i think the significant conferences are NIPS (http://www.nips.cc/), ICML, UAI (e.g., http://event.cwi.nl/uai2010/), and AISTATS (http://www.aistats.org/).

dododoon Every "best paper" from Computer Science conferences since 1996

the title is wrong. it's not every "best paper": notably all best papers before 2005 are missing for ICML.

in terms of metrics of research quality, not all conferences are equal, and "best" papers are often selected by a relatively small group of people whose decision isn't really validated too much.

dododoon A Censorship-Resistant Web

this reminds me of ross anderson's eternity service from ~1997:

  The Internet was designed to provide a communications channel that is as  
  resistant to denial of service attacks as human ingenuity can make it. In this note,
  we propose the construction of a storage medium with similar properties. The 
  basic idea is to use redundancy and scattering techniques to replicate data
  across a large set of machines (such as the Internet), and add anonymity 
  mechanisms to drive up the cost of selective service denial attacks. The 
  detailed design of this service is an interesting scientific problem, and is not 
  merely academic: the service may be vital in safeguarding individual rights 
  against new threats posed by the spread of electronic publishing.

http://www.cl.cam.ac.uk/~rja14/eternity/eternity.html

an implementation of which can be found here (also from similar time):

http://www.cypherspace.org/eternity/

dododoon C++ Streams & Typedefs: Be Charful

haskell does support it.

c style:

  type Foo = Int

pascal style:

  newtype Foo = Foo Int

dododoon Kinect sex game will not become reality, Microsoft says

sex != porn. hopefully i'm not here today because of porn.

dododoon Tim Berners-Lee says Facebook is a trap

w3c should create open standards for social networks.

i don't think these companies are going to do it, and this seems exactly what w3c is meant to do. so why not? seems more productive.

dododoon [dead]

i wonder how the TSA employees feel about doing this?

what kind of person would apply for such a job, knowing this was one of the duties?

how is this not state-sponsored sexual assault?

... definitely trying to avoid the usa for the foreseeable future. too fucked up.

dododoon Facebook Finds A New Way To Liberate Your Gmail Contact Data

that's revenue, not profit.

dododoon Facebook Finds A New Way To Liberate Your Gmail Contact Data

they have ~500 million users[0]. that's ~$4/user/year.

[0] http://www.facebook.com/press/info.php?statistics

dododoon Facebook Finds A New Way To Liberate Your Gmail Contact Data

or "facebook won't let you liberate your facebook data to alternate providers".

i don't understand why people tolerate facebook. they don't seem deserving of the trust people give them. first opt-in/opt-out privacy issues, now this... plus it doesn't seem like they're really trying to make money yet. i suppose this is what the initial "exclusivity" of the facebook brand got them: loyalty without needing trust.