alextp

26 ms·

241.
scale.metaoptimize - A StackOverflow-like site about high scalability and NoSQL(scale.metaoptimize.com)
74 points by alextp 16y ago|22 comments
242.
by alextp 16y ago
Really thank you for that. I was trying to use quora for this purpose, but it's not specific enough. As a machine learning phd student from somewhere far from most good research centers (I'm in brazil, and how many brazillian ML papers have
243.
by alextp 16y ago
I have an agreement with my thesis advisor in which I will send him many emails as I think through an issue, and he doesn't have to read them carefully (or at all). I will clearly mark the parts that he needs to answer (like the time of the
244.
by alextp 16y ago
Yes, it is the case here that a bayesian approach can lead to a previous empirical formula falling out. What I was saying as well is that this regularization + test set approach is also valid (and sometimes slightly more or less general tha
245.
by alextp 16y ago
Yes, but the formulation I showed you is equivalent to a bayesian prior. For example, if you want to learn a weight vector w that gives high likelihood to the data and has a gaussian prior with 0 mean and C identity covariance, the MAP answ
246.
by alextp 16y ago
Peer review is usually by scientist peers, and not professional peers. I doubt that most academic researchers in medicine were prescribing anything, really.
247.
by alextp 16y ago
Yes, that is the main pitfall with machine learning in general. Usually you do two things to mitigate it (but they don't fix it entirely): regularization and a test set. Regularization is some means of complexity control. You switch form tr
248.
by alextp 16y ago
Except verifying termination with a bound in the seconds of computation can be done in O(n), where n is the maximum number of seconds allowed (you can see that trivially by just waiting n seconds and killing the program if it doesn't termin
249.
by alextp 16y ago
Does he have a blog or any such thing? Is this text on archive.org?
250.
by alextp 16y ago
I do worry about this, when I consider working a few years at banks after graduating to save up some fuck-you money.
251.
by alextp 16y ago
An interesting thing posterous could to is send the user a (daily? weekly?) email "reminding" him of the blog, and making it so that just replies to that email count as posts. This lets them even change the GUID for each user if they think
252.
by alextp 16y ago
I'm interested too: alexandre.tp at gmail
253.
by alextp 16y ago
Probably this has something to do with wiretapping telephones or something like that.
254.
by alextp 16y ago
As long as they are integer keys, that is.
255.
by alextp 16y ago
I think he meant the musician formerly known as Prince for this one.
256.
by alextp 16y ago
But they must surely be doing something fancier than this naive-bayes-style model, otherwise they'd have no use for a roomful of supercomputers.
257.
by alextp 16y ago
Not only that but you can probably parse a jeopardy "answer" into a series of roughly independent clauses, and then try to predict classes that rank high in these clauses. For example, in the "answer" "This action flick starring Roy Scheid
258.
by alextp 16y ago
> 1. Poor grammar and bad writing are often a sign of poor comprehension. This is true, but I'd add "at least be the poor reader".
259.
by alextp 16y ago
You can search for local linux/debian user groups.
260.
by alextp 16y ago
This is actually true. Since moving the last time, I've found that a weekly trip to the farmer's market a few blocks away from home, with a stop in the butcher on the way back, solves 90% of my grocery needs. Beer, cleaning supplies, and gr
261.
by alextp 16y ago
And how exactly would the many inevitable civillian deaths and the gruelling guerrilla that are certainly going to happen if the US invades north korea help anything? It's partly because of war-mongering articles like this from the NYT that
262.
by alextp 16y ago
LastFM and Spotify should be compared with having your videoclip on MTV or your song on radio, not with selling albums.
263.
by alextp 16y ago
That could be an interesting startup; maybe coupled with a to-do list or a calendar or some such productivity tool. Maybe giving the users some "free" perk randomly whenever they advanced upon some self-set goal could work.
264.
by alextp 16y ago
I think it might have something to do with translating. These "confusing" sets of words are translated as completely different things in portuguese. On the other hand, my high school teacher usually used "why" and "because" to explain the d
265.
by alextp 16y ago
Oddly enough I think some kinds of foreign learners end up _more_ worried about some aspects of written english than native speakers. Things like they're/there/their, affect/effect, it's/its etc are more likely to deceive someone who learne
266.
by alextp 16y ago
No memory usage data? That's usually important for heavy-duty optimization.
267.
by alextp 16y ago
When talking about complexity you usually have to drop the "turing machine as a metaphor for everything" idea very fast. In complexity theory you're interested in how fast will some different variations of computing machines (that might not
268.
by alextp 16y ago
Although it can be solved. Let me preface this by saying that I usually dislike privatization of public utilities. When I moved from Salvador to São Paulo, one of the biggest differences I saw was in the cartórios (civil registry offices, I
269.
by alextp 16y ago
She's apparently saying the most trivial statement possible about bribes: if nothing works the way it should but something will work if you bribe someone, then bribing these people will make these things work. Well, duh.
270.
by alextp 16y ago
I think this is a valid use case of spam filters. I have trained more than one to detect my father's powerpoint emails and bad chain-mail jokes and separate them from his personal messages that I actually want to read.

More ›