HNHacker News
TopNewBestAskShowJobs

luchak

267 karma · joined February 3, 2009

email: luchak@luchak.net web: luchak.net
submissionscomments
luchak··on Functional Fluid Dynamics in Clojure
The link only uses Euler integration for the diffusion (advection is semi-Lagrangian), and in graphics you never use very much diffusion anyway, because highly diffusive fluids are boring. In fact, when using this particular technique, it's not uncommon to set diffusion to 0, since the multilinear interpolation used in the advection step tends to introduce a fair amount of diffusion.
luchak··on Functional Fluid Dynamics in Clojure
Honestly, in a lot of cases language performance doesn't matter as long as you have a good numerics library. For example, here's one implementation of the projection step in my current project, which does fluid simulation on unstructured tetrahedral meshes and relies heavily on SciPy's sparse matrices:

  def __init__(self, mesh):
    div_matrix = op_mesh.n_to_nm1_boundary.T * op_mesh.F_matrix
    boundary_matrix = op_mesh.F_matrix - op_mesh.FB_matrix
    self.C_matrix = sparse.vstack((div_matrix, boundary_matrix))
    self.CCT_matrix = self.C_matrix * self.C_matrix.T

  def Project(self, velocities):
    lam = linalg.cg(-self.CCT_matrix, self.C_matrix * velocities)[0]
    return velocities + self.C_matrix.T * lam
It's pretty simple (constrained least squares optimization) and it's missing some stuff (preconditioner), but it runs fast enough (~1 minute/frame for > 1M tetrahedra), and only required minimal testing: it's all built up from operators I was already using elsewhere. Since it's the fourth or fifth projection method I've tried for this code, that makes a big difference. The fact that I'm using Python is almost immaterial: all of the "interesting" code is library calls.
luchak··on An appeal for unremarkable women
From Wikipedia:

"Sexual objectification refers to the practice of regarding or treating another person merely as an instrument (object) towards the person's sexual pleasure."

I don't know of anyone who wants this in the context of work. Even if they're, say, attracted to someone else on the team: they're still not interested in being reduced to only the sexual aspect of their being.

luchak··on An appeal for unremarkable women
No, no, no. If you say "okay, we won't do that" and mean it, you've demonstrated respect for her, and that her opinions and preferences are valued. You've included her in the group.

And for your second paragraph: saying "no, it's not like that" is marginalizing because it demonstrates an unwillingness to take the woman seriously. If she says, "I didn't like that, it made me feel excluded from the group", and you say, "well, we didn't mean it that way": you are continuing to exclude her. The message communicated is, "your opinions aren't important". If the presentation went so far as to be objectifying, you're saying "well, I didn't mean for it to make you feel objectified, so you shouldn't feel that way" -- which indicates a flagrant lack of respect.

For emphasis: I do not know a single female engineer (not a single engineer, actually) who wants to be sexually objectified by his or her peers. Not one. They want to be respected as competent engineers, as colleagues, as friends.

luchak··on An appeal for unremarkable women
It's not like somebody's sitting there saying, "hey, it's a woman -- I should make her miserable." It's more like there's a category of crap that ends up shoved at you if you're female, but not if you're male. There's another category that men deal with too, but it's much more potent when you're a small minority of the population. And, I mean, come on, some of these things: the attractive=stupid stereotype, the facile labeling of some women as irrational, being mistaken for secretaries or assumed to be non-technical staff ... these play upon stereotypes and cultural norms that apply specifically to women.

In other words: because most of these things don't apply to men. And because, when women try to explain what's happening, one of the most frequent reactions they get from men is, "what are you talking about?" (Or the related, "oh, but that happens to me too," followed by a spurious equivalence.)

So, look, if you really want to get into the details, there's a limit to what I can tell you. I'm male, so I'm not speaking from my own experience. If you're genuinely curious, ask some female engineers. Not all of them will feel the way I've described, but I've talked to enough to know that I'm not totally off-base. The main thing is, if you want to know, you have to be willing to actually listen and accept that they have had experiences different from yours. Even if you disagree with them, try to hold off on going into debate mode; they've probably done that too many times before and are tired of it.

luchak··on An appeal for unremarkable women
Yes, these things are important to both men and women. But the female engineers I know make it a higher priority in their job searches because they've been burned this way before, in work and in school -- specifically because they are women. I don't know any male engineers who feel that they've been in situations, in the context of engineering, where they were systematically dismissed because they were men.

I mean, you hardly need a locker room atmosphere to turn women away. Plenty of little things can contribute. Maybe you have a guy who constantly fawns over one particular coworker. Maybe the team has grown to expect that the women will bring the bagels on Friday mornings. Maybe the labels "emotional" or "too sensitive" stick a little too easily. Maybe t-shirts at work events only tend to show up in men's M-XL. Maybe another group tends to explain things to the new female engineer in very small words, implicitly not acknowledging her as a peer. Whatever.

Here's the thing: you need to be sure that the women in your workplace are comfortable telling you these things. As a man, it can be hard to realize that there are problems when the work environment is 90% men. Why? Because you're not on the receiving end of all the bullshit, most of the stereotypes and cultural norms work in your favor, and most of the people you know are the same way. Very few men are thinking, "man, I wish I could make women feel unvalued." Most of the toxic behavior happens out of apathy or ignorance.

But, even then, most little problems, in isolation, are just that: little problems in search of solutions. Unfortunately, when women do speak up, what they often hear is, "I don't see the problem. That's never happened to me. I think you're overreacting." That response is, I would venture, why having another woman on the team is reassuring: because she probably would see the problem, it probably has happened to her. That response is why women in engineering want to know they'll have someone (male or female) willing to back them up, or at least to listen: because they're used to having to go it alone.

luchak··on An appeal for unremarkable women
From a recent email exchange with a female software engineer (heavily paraphrased):

Get to know some female engineers, if you don't already. Not necessarily as potential hires, but as professional contacts. At that point, they may decide they're interested in the jobs you have, or they may know women who are. If they refer female candidates to you, their implicit endorsement will help your credibility. It will be very hard to hire female engineers if you don't know any.

luchak··on An appeal for unremarkable women
I tried to address this in the last paragraph of my post, but it's a hard problem, and not one I know how to solve. Some of it, undoubtedly, also comes down to timing, location, and luck.

But you're right, I should have a better answer to this question. I should discuss it with my friends sometime and try to get a handle on it.

luchak··on An appeal for unremarkable women
If you can't hire any female engineers, part of the reason may be that you don't already have any. This can be a huge red flag. Basically, most women applying for the job want to have some assurance that they're going to be respected and their concerns taken seriously, and the best way to demonstrate that this is true is by having some women on your team. Otherwise, the question just sits out there: why don't you have any? Is it coincidence, or is there something wrong with your culture?

(Disclaimer: I'm a man, but I know a lot of female engineers. I obviously can't speak for all of them, but I've noticed some trends.)

Many women in engineering are wary of situations in which they won't be taken seriously.

For example: say there's one guy on your team who tends to patronize women, especially ones he finds attractive. He might not realize he does this. You have a female engineer who's getting fed up with this behavior, and tries to gently (because she's experienced this before) point this out. He gets defensive -- maybe he says she IS different than the other engineers, so it's only natural, maybe he just gets angry, whatever -- but, whatever he does, it's not an honest effort to improve. Later, she tries to bring this up with someone else on the team, look for help, but to everyone else he's always been a nice guy, so why is she making trouble?

Or let's say someone illustrates a presentation with some unfortunate images or metaphors. (Where have we seen this before?) Nothing flagrantly unconscionable, but, you know, kind of skirting the edge of acceptability. This makes a female engineer uncomfortable -- the presentation marginalizes her, makes her wonder if the team sees her as a colleague or an object of desire. Now, she knows it's going to end badly, but she mentions that it made her uncomfortable, and asks that everyone please try to avoid that mode of presentation in the future. And suddenly everyone on the team is explaining, no, he didn't mean in THAT way, the presentation wasn't REALLY about her -- well, if you hadn't marginalized her before, now you definitely have.

It doesn't have to be that way. But in some places, it is. I've talked to women looking for software engineering jobs, and this is definitely something that gets considered. Are other women willing to work there? Willing to stay? Will they have anyone to back them up when shit happens? Because when the answer is "no", life sucks.

That said, I don't know how you hire the first one. But you've got to be aware of these things. You've got to make completely sure that you look like someplace where women will be respected and taken seriously, not marginalized and undermined. Because, if a good female candidate gets a whiff of that, she's probably gone.

luchak··on Show HN: Review my startup: Quora, a topic based question and answer site
In approximately inverse order of importance (sorry about that):

* Misc.:

- Don't like using my Facebook account.

* UI:

- Always find myself hunting for the "follow" button.

- Front page is a visually undifferentiated mishmash of 20ish different questions, most of which contain no more than a few sentences of content. Also: do we really need number of answers, number of topics, follow, star, time, topics (selected arbitrarily?), profile picture, answer snippet, answerer name, answerer description, and answer votes for every single one of these questions? You're making me sort through way too much stuff to decide whether I'm interested in the question or not.

- The options menu is cryptic and poorly located. Ex.: If I want to answer something anonymously, why do I go hunting at the right-hand side of the screen by all the navigational / related links stuff? Put that by the text box.

* Community:

- The personality types on the site are not substantially different from people I already know. Most of the questions (both from my feed and via "All Questions") drip Silicon Valley techie. I'd love to see intelligent questions/answers from people who have lives very different from mine. As things stand, I'm not sure whether, if I were to ask a question, I'd be getting a range of opinions or an echo chamber.

- Answers tend pithy. On the other hand, some of the questions encourage that. Is there some way to see only questions/answers with meat?

Summary: Right now I feel like Quora has a lot of features (too many features) and not enough content, even in what look to be active topics. I don't know how you get people to invest in writing high-quality answers, but that's the question you really need to be focusing on. I would say that you should try to to cultivate a sense of sitewide community, but the siloing of the site into a bunch of semi-intersecting topics could make that tricky.

But then, the only Q&A site I've ever really liked has been Ask Metafilter, so maybe I'm looking for the wrong thing here -- but I don't think so, since, for me, it's the first site that comes to mind when I think of your mission. (Put together a better-indexed version of Ask Metafilter and you'll have a winner in my book.) If I need to ask a short question that can be answered quickly, I'll consider Quora. If I'm looking for significant thought, longer meditations, a diverse range of opinions, or interesting browsing material, though, I'll head elsewhere.

luchak··on Imagine There's No YouTube
That bar analogy is sounding awfully tenuous by this point. Was YouTube paying people to upload the video?
luchak··on Image Space Photon Mapping: The insanely great near future of computer graphics
Neat! I can't say I'm familiar with the paper, but I'll check it out sometime.
luchak··on Image Space Photon Mapping: The insanely great near future of computer graphics
This is definitely very cool. I was actually fortunate enough to attend a presentation that the first author gave just a couple of days ago. It was a good talk and the slides were substantially based on the ones from the project page, so I'd certainly recommend checking those out.

A couple of notes from the talk (paraphrased):

- Image-space photon mapping solves the same problem as traditional photon mapping; it does not introduce any extra simplifying assumptions. The performance improvements come primarily from (a) acceleration of the first bounce using rasterization, and (b) the improved radiance sampling technique yielding good results with 100x fewer photons than traditional methods.

- The slides that I saw included some extra material, including a list of other common numerical problems that can be relatively easily formulated as special cases of the rendering equation. Since image-space photon mapping produces randomized solutions to the full rendering equation, there may be extensions from this work to other important problems. (He drew an analogy between convolution and rendering a scene computing only the first bounce.)

- Q: It's not clear to me exactly how to convert some of the other numerical problems you mentioned into scenes for rendering. Which of your contributions do you think might be most valuable to people working on these problems? A: The radiance estimate. I think the lesson to draw from that is, no matter what your problem is, it's a good idea to try using fewer samples and extracting more information from each one.

- Q: Can you handle volumetric effects or sub-surface scattering? A: No, we assume that all scattering occurs at an object surface.

- Q: Why are you attempting to achieve physical accuracy? Studies have shown that people are really bad at judging whether the scene they're looking at is physically accurate. A: You don't want your artists to have to understand the guts of your rendering techniques. When you start using hacks to capture only a few perceptually relevant effects, the result can be fragile, and as soon as the artists change anything the scene looks terrible. If your renderer is physically correct, though, everything just works.

EDIT: Made point (a) in the first note a little more precise.

luchak··on Nearby "Super Earth" May Have Oceans, Thick Atmosphere
I'm not so sure, given the article as written. Comparing Earth and Jupiter:

1. made up of about three-quarters liquid water

False for both. Possibly true for Earth if you're talking about surface area, but that's a pretty generous interpretation of the article as written.

2. with a solid core of iron and nickel

Likely true for both. ("At the center of [Jupiter's] rocky core is probably a metallic ball of iron and nickel, just like Earth's core." http://www.spacedaily.com/reports/Jupiter_Has_Large_Rocky_Co...)

3. an atmosphere of hydrogen and helium

True for Jupiter, not for Earth.

That's 2 of 3 for Jupiter, 1 of 3 for Earth (or 2 of 3 for Earth if you're being generous). Mars sounds a lot more like Earth than GJ 1214b does, based on the article.

GJ 1214b might be more like Earth than any other known exoplanet, but given the murky reporting it's hard to tell.

luchak··on Nearby "Super Earth" May Have Oceans, Thick Atmosphere
"Based on its density, scientists think GJ 1214b is made up of about three-quarters liquid water with a solid core of iron and nickel and an atmosphere of hydrogen and helium—not unlike Earth."

what

This is either terrible writing or terrible science (on the author's part, not the scientists', I would imagine). About 70% of the Earth's surface is water, but only a fraction of a percent of Earth's mass is water. (http://bit.ly/7Gfk8V, total water volume from http://en.wikipedia.org/wiki/Water#Water_on_Earth). And the Earth's atmosphere is definitely not primarily hydrogen and helium.

luchak··on Microsoft Statement Regarding MSN China Joint Venture’s Juku Feature
What?

The Chinese government has issues with sites containing user-generated content. Neither Microsoft nor anyone else can resolve these issues just by throwing money at the problem. Getting Plurk back into China would likely involve making substantial changes to how the site is run: for example, demonstrating that they can and will take down or block posts at the request of the Chinese government within a specific amount of time after the request is made.

luchak··on Microsoft rips off Plurk
Something like this did happen at Google China a couple of years ago: they released a Chinese IME that contained a dictionary taken from Sohu. Sohu demanded a public apology, which Google issued a couple of days later, along with a version that did not contain the stolen list.

http://www.webmonkey.com/blog/Google_Apologizes_For_Poaching... http://bit.ly/5zIVYl

Edit: for what it's worth, I personally agree with the theory that this was probably done by a local team that was under a lot of pressure but wasn't getting enough oversight.

luchak··on Tis the season for Grad School Advice
The guide that basically walked me through the entire application process was also written by a CMU professor. It's a much drier read, but, if you're applying to Ph.D. programs in computer science, it's gold.

http://www.cs.cmu.edu/~harchol/gradschooltalk.pdf

luchak··on A successful academic faces lifelong debt
http://www.insidehighered.com/news/2008/04/14/aaup might be a good starting point ... looks like the average for an assistant professor is around 60k, but Dartmouth is likely to be above that.
luchak··on Why Go to Graduate School and Getting into the Program of your Dreams
Graduate school has a number of extremely important advantages over independent research, at least when you're getting started: 1. Mentorship. 2. Easy access to the rest of the research community. 3. Access to resources: libraries, journal subscriptions, and, most importantly, funding.
luchak··on Official Google Research Blog: Machine Learning with Quantum Algorithms
As Wikipedia says, in a quantum computer, you have a quantum register (a bunch of entangled qubits) that holds the result of your computation. The contents of this register can be visualized as a high-dimensional vector of length 1 -- think of this as a line with one endpoint at the origin and the other restricted to the radius-1 hypersphere around the origin. Each axis of the space that this sphere lives in corresponds to one classical value of the register -- in an 8-qubit register, one axis corresponds to 00000000, another corresponds to 00000001, etc.

The real problem you solve with Grover's algorithm is this: you have a black box (usually called an "Oracle operator", but I think the term "Oracle" in this case is misleading) that negates the value of your register along exactly one axis (i.e., if the coordinate on the 00010111 axis is .2 - .1i, it becomes -.2+.1i). You want to figure out which of the 2^n axes in the n-qubit register your black box negates.

Unfortunately, you don't have the luxury of picking the vector that has a coordinate of 1/sqrt(2^n) in each direction, and observing what the black box does to that after a single application, since when you observe your quantum register you always observe a classical state. In contrast, one solution that would work would be to solve the problem the classical-computing way: march through each of the 2^n axes until you find the one you're looking for.

Grover's algorithm gives you a better way. What Grover's algorithm does is repeatedly apply an operation that nudges your register towards the desired state -- basically applying a small rotation. After some number of iterations, the vector representing your register should be pointing in almost the same direction as the vector representing the state you're searching for. Then you can measure the contents of the register and, with high probability, get the answer you were searching for. It turns out you only need O(sqrt(2^n)) queries (rotations), so it's a substantial improvement over the classical case.

I've probably oversimplified a few things, but hopefully this makes things clearer than they were before.

luchak··on Why Go to Graduate School and Getting into the Program of your Dreams
Hm. So what's the alternative? I've tried to do research in a corporate environment, and am now in grad school: in my experience, it seems like getting a Ph.D. has a lot to offer in terms of teaching you how to do good research. I've learned a lot more about both my area and about the practice of research in general in graduate school than I did when I was working in a corporate research group.

I'm not convinced that working at a smaller company (or, since this is HN, starting one) would have helped, either. In fact, I'd guess would have been even worse. It would have been harder for me to gain access to advice and mentorship, there would have been more distractions, and the pressure to stay relevant to the goals of the business would have been even more intense.

luchak··on Why Go to Graduate School and Getting into the Program of your Dreams
Hm ... I'm puzzled by your last sentence. Why do you say that?
luchak··on Ask HN: Review my startup, GetPlaylists
As other people have said, the stock-photo girl has got to go. It makes me feel like you're trying to sell me insurance or tooth whitener or a mortgage or something.

My main issue after poking around the site for a couple minutes is that I still can't tell what you offer that I can't already get from iTunes and sites like Pandora/Last.fm/etc. The iTunes "Genius" feature works well enough for assembling quick 'n dirty playlists out of my current music collection. If I want to share my music tastes with my friends, I'm already on last.fm. If I want to share a specific playlist with my friends, I'll burn them a CD or something, since their collections don't overlap with mine much.

Maybe I'm not your target audience, but, regardless, here's what I'd like to see: really interesting curated playlists. I can't find any on the site right now. The playlists I can find now are all very, well, functional, but they're not compelling. Ideally these playlists would be more like mixtapes than unordered jumbles of songs. It seems like it wouldn't be too much trouble (or too expensive) to get some good online or college radio DJs to assemble some quality content for you. You could also just use show track lists as playlists, but the results wouldn't be quite as nice.

(It'd also be neat to get suggestions for playlists that contain mostly, but not entirely, music I already own.)

That's the smallest delta from the current site that I would find motivating. I'd argue, though, that there's still nobody who really does music discovery "right", so there's a lot of room for innovation. I'm glad to see you're exploring the space.

luchak··on Why Go to Graduate School and Getting into the Program of your Dreams
I know someone who wanted a Master's degree, but applied to Ph.D. programs so he could get it paid for. Things didn't go exactly as planned, and now he's a computer science professor.

He seems pretty happy with the outcome, though.

(I don't really recommend doing this, but if you're strongly leaning Ph.D. anyway and you have some research experience, you should probably just go for it.)

luchak··on Review our webapp- The Uncertain Future (future world modeling)
Really cool. Would be interesting to break some of these things down further, into easier-to-estimate probabilities -- probably using some sort of more detailed graphical model. Seeing some sort of consensus view of the future according to your users would be neat, too (especially if it were broken down by demographic).

The interface is a little quirky. I sometimes found it difficult to determine how to set means / deviations.

luchak··on [dead]
I'm okay with that, I just don't like the affect they effect.
luchak··on Feeback on my new site, please...
I still don't get it. You've described an abstract issue: universities have a fair amount of power to control what messages about them are disseminated. But what are the use cases? Why would I want to visit your site? Give us a couple of specific examples, preferably in the form of stories.

Also, I would say that the interface needs to be revamped. Right now it's just confusing: it looks like a blog, none of the posts are even marginally relevant to me, and I have no idea how to get to any content I want to see -- or even what that content might be.

I could see a case for the existence of some sort of university-targeted Stack Overflow-like site. Basically public Q&A boards for particular universities, structured by topic, and with a good method for pulling useful content to the top. Maybe this already exists and I don't know about it, or maybe it's not what you're going for, but I at least know why I would use it.

luchak··on XKCD Explained, One Strip at a Time
I feel like this blog is an optical illusion where, if I stare long enough, I'll see John Campbell.

Except I'm still not seeing it.

luchak··on Educational: Worlds Worst Website
This would be more educational as a collection of pages that each get one or two things wrong.
← PreviousPage 2 of 3Next →