HNHacker News
TopNewBestAskShowJobs

maldeh

519 karma · joined September 6, 2013

submissionscomments
maldeh··on Oculus Joins Facebook
I see the absence of overlap as a good thing. It means it's highly unlikely that Facebook threw down $2 billion for some gimmicky social integration that a tiny sliver of their userbase could harness.

The Oculus team is cutting-edge hardware along a completely different vertical, orthogonal to facebook's core. Despite their size they have the talent and expertise to create that entire market by themselves, provided adequate funding. Facebook does not have to interfere at all, beyond writing the cheques. Over three years, Facebook could have itself a new gaming product to compete with Sony and XBox... in five, it might have an innovative consumer device to take on Apple. The conglomerate model typically has better odds of long-term survival - what happens when people grow tired of social in its current form (primarily content curation)? Then facebook's flagship product would be obsolete.

maldeh··on Top Algorithms in Data Mining (2008) [pdf]
Kevin Murphy's treatment of Kernel Methods in "Machine Learning - A Probabilistic Perspective" offers a take on SVMs that I don't often find in most discussions. He makes the case that SVMs are generally rather unusual from a Bayesian point of view. Apropos to other sparse vector machines, Murphy opines:

"Note that SVMs are very unnatural from a probabilistic point of view. First, they encode sparsity in the loss function rather than the prior. Second, they encode kernels by using an algorithmic trick, rather then being an explicit part of the model. Finally, SVMs do not result in probabilistic outputs, which causes various difficulties, especially in the multi-class classification setting.

"It is possible to obtain sparse, probabilistic, multi-class kernel-based classifiers, which work as well or better than SVMs, using techniques such as L1VM or RVM... However, we include a discussion of SVMs, despite their non-probabilistic nature, for two main reasons. First, they are very popular and widely used, so all students of machine learning should know about them. Second, they have some computational advantages over probabilistic methods in the structured output case... [specifically, SSVMs]

...

"The conclusion is as follows: if speed matters, use an RVM, but if well-calibrated probabilistic output matters (e.g., for active learning or control problems), use a GP. The only circumstances under which using an SVM seems sensible is the structured output case, where likelihood-based methods can be slow. (We attribute the enormous popularity of SVMs not to their superiority, but to ignorance of the alternatives, and also to the lack of high quality software implementing the alternatives.)"

----

the tl;dr is that SVMs are an important tool for their historical impact and popularity, but the modern data scientist has a range of cookie-cutter algorithms to choose from that could possibly be much more useful.

maldeh··on Why Isn't the Fourth Amendment Classified as Top Secret?
From a policy standpoint, I like to think of the Senate Intelligence Committee as a good-faith attempt at providing "sufficient statistics" of otherwise sensitive data -- having trusted elected officials peruse classified information, collate it into the essential set of decisions the public might formulate should they have full access, then let the public debate proceed along the transformed decision space.

For the outcomes to be truly unbiased, a number of assumptions should be enforced, that are being insufficiently upheld: - the sample of the senators in the committee should to be representative of the people's interests. The 8/7 member split between parties sounds nice and non-partisan but might be biased in favour of the minority. Also, nomination to the post depends on party favour. - a mechanism guaranteeing that the intelligence committee obtains ALL pertinent documents sans curation by the agencies. It's certainly ridiculous that there's even the possibility the CIA could claim that the Intelligence Committee obtained some documents illegally. - finally, a notion of weighting actions appropriately. The intelligence committee is not in the business of taking action in itself, but making recommendations for the rest of congress to act on (subject to weights based on their political stances and the requirements of the situation). These weights are unfortunately being re-normalized to near even odds along purely political lines in the senate and in the media.

While the representational biases can be repaired somewhat with appropriate procedure, the final problem of retaining the right weights for actions doesn't seem solvable so long as the first amendment is around. One can't silence Fox/MSNBC pundits constantly trying to put their spins on every debate, and it's difficult to correctly evaluate these solutions unless the public can actually gauge the adequacy of each approach against all the data.

tl;dr: IMO the ability to take optimal rational decisions via proxy opinions from the intelligence committee is severely mitigated by biases inherent in the institution. So long as the committee model persists it may be possible to mitigate these biases but impossible to completely eliminate them.

maldeh··on Go Ahead, Let Your Kids Fail
to be fair, she DID take the trouble to approach a science teacher on a subject she didn't know much about, instead of making a grading decision with insufficient knowledge.
maldeh··on A 33-Year-Old NPR Story Convinced Me Google Glass Will Stop Looking So Dorky
CCTVs are accompanied with the assumption that some authoritarian figure is responsible for the recordings, and will follow reasonable and legal guidelines to distributing the feed. Even IP Cameras with live streams to the internet are subject to township / state / federal laws and cannot record near bathrooms, hotel rooms or anywhere where there is a reasonable expectation of privacy.

A closer analogy to Google Glass might be handycams and smartphones, and these face plenty of pushback even without the always-on criterion.

maldeh··on Start unique and drop the “the” – some thoughts on URLs
That's fair, but is it really so much work to enter three extra keys on a url? That doesn't strike me as a convincing enough reason not to attempt a brand change, if it otherwise makes sense. (Unless there are many frustrated mobile users somewhere who've decided they've put up with long urls far enough!)

Using an article such as "the" in a brand typically indicates a desire to lay stake on a common noun as a proper noun, or at least the definitive instance of the noun. It often doesn't make sense when your brand is obviously a proper noun or otherwise unique already, but I suppose there's plenty enough newspapers having a variant of "Guardian" in their titles to warrant a push to definitive-hood.

Edit: A friend also pointed out that "guardian.com" is already taken by another established organization. I think that trumps my silly grammatical analysis :P

maldeh··on I Am An Object Of Internet Ridicule, Ask Me Anything
the whole problem likely stems from this very bothersome definition of a hipster. At a baseline, a hipster is simply "a person who follows trends and fashions in an ironic way". Different people have these other nuances and qualifiers that are applied inconsistently... "they're not a hipster unless they're also shallow and facile"... "hipsters are attention seekers"... "a hipster is somebody who tries to be one, until it stops being cool" and so on.

What irks me about these various definitions is that it's as if people are having an academic debate about which pejorative or negative stereotype deserves to be christened a "hipster", when there's already a very diverse and complex community in existence that self-identifies as such. The only context I see such a debate making sense is when people have a perverse need to somehow preserve their ire and rage against the word "hipster" after being confronted with wholesome, happy counter-examples.

This poor bloke is a guy whom a lot of regular people would quickly slot down as a hipster, just from his looks and habits. He probably self-identifies as one too (because, hey, he likes the same music and dresses the same and it's always nice to have a sense of community with like minds). And there ought not be anything wrong with it. Yet there's an active hipster-hate community that's not going to drill down into his benign motivations before shredding him up.

So what appears to be the way forward here? Collectively agreeing that it's high time to dial back the rhetoric on hipsters? Or telling the guy who thinks he's a hipster that he's not really a hipster because hipsters are supposed to be jerkfaces?

← PreviousPage 5 of 5