does ireland have a legal concept similar to common carrier in the u.s.? i'm not a lawyer, much less an expert on the irish legal system, but it would seem to me that this case could only work in a country where common carrier laws are either non-existent or very weak. if ireland does have something like common carrier that would cover eircom then a win appears to essentially invalidate common carriers and make any isp that sends traffic through ireland potentially liable, even if both ends of the infringing connection are outside of irish jurisdiction.
haar, 1910, is the earliest example of a wavelet in a paper. however, you could argue that the core concept (projection onto a functional basis) goes back to fourier, with haar's real contribution being finding a particular functional basis that obeys the rules associated with wavelets. to the best of my knowledge, most of the wavelet math was worked out relatively recently by mallat, daubechies, etc. starting in the '80s
i don't think that one anecdotal data point really means that hiring people with proven records of delivering quality results on time and training them is a bad idea so much as it underscores the importance of not hiring losers. quite frankly, if it takes someone much more than a month to three months to develop a productive level of proficiency in a new language, you're dealing with a dud and it's time to let them go.
i think they were hoping that you'd see the patterns in the exercises and figure it out for yourself.
the tricks that i used most frequently were:
when you see tan(x) by itself, you probably want to rewrite it as -(-sin(x)/cos(x)).
keep a look out for things that look like trig identities particularly when dealing with rational expressions (e.g (tan^2(x)+1)/csc(x) = sec(x)/csc(x) = (1/cos(x))/(1/sin(x)) = cot(x) = cos(x)/sin(x) ==> du/u) -- your goal is to use those identities to bring things into a form that you know how to deal with.
you want usually want to pull things like cos^n(x)sin^(n+1)(x) out by themselves (==> 1/(n+1)u^(n+1)du) by themselves. if you've got unbalanced powers of sin and cos, you probably want to try a power reduction identity to reduce one of them to something a bit more manageable or see if you can do something with the product rule.
i can't speak for your t.a.s, but my experience was this: i worked my ass off in labs trying to cover the gap between what the lectures had covered and what i thought the students needed to know to do the projects and homeworks. as a result, the people who showed up regularly to labs never scored less than a B on the projects and programming assignments. people who didn't come to lab were all over the map. likewise, i worked my ass off to be prepared for office hours. the students who came to my office hours regularly all passed and did a lot better on average than the people in the class who didn't come. there's no t.a. that can force you to come to office hours or make you ask questions like, "if you were grading a problem like this on the exam, what do you think the professor will be looking for when it gets graded".
for what it's worth, if you're memorizing rather than learning the small handful of tricks that you need to do the trig integrals, you've missed the point.
on the other hand, you can occasionally luck out like i did --- i teach robots how to see and the gov't pays me handsomely for it. i can't really thing of a more awesome job:-) getting those awesome jobs requires a combination of technical mastery that goes well beyond what an o'reilly book would cover, excellent soft skills (communication, interpersonal, leadership), and the ability to self-promote. that last one is, i think, extraordinarily unnatural for most technical people but tends to come more easily to business types.
your point about the problem starting in k-12 education is well taken, and i certainly agree.
i agree that the math core and science can be a major stumbling block when taught poorly --- lord knows i've had more than my fair share of poor math and science profs. that said, there's a major disconnect between how high school in the us fails, miserably, to prepare students for college level math and the fact that math at a university level actually is a lot more than quality time with your ti-89. there are a lot of people who approach university math with a high school math mindset, people who are used to sailing through with no intellectual challenge because teachers are teaching to the test with cookbook answers. my first university math class was a real wake-up call for me in this regard, although i was lucky enough to wake up early enough that i didn't fail spectacularly.
as a t.a. (so someone who's been on the other side), i had several students come to fight with me when they got their papers back because they felt they deserved more points because their answer and the real answer had some of the same symbols in the same places. it's right, or it's wrong, and while i did what i could to give as much credit as i could for the level of mastery they had demonstrated, there was no way that i was going to give them more credit than they deserved. most of them deserved to fail.
i suspect that this is what you're running into: the combination of people actually expecting you to perform well (i.e. at a far higher level than what high school expected) and no longer being able to slide by without putting effort into the classes.
honestly, i'm surprised you're getting away with so little math and science. we were required to do calc1-3, discrete math 1, linear algebra 1, prob and stats for engineers (what would have been two semesters in the math department crammed into one, with a sadist of a prof), and numerical methods. we also had 3 semester of lab science, including at least 1 sequence (i did chem1-2 and bio1). this doesn't cover the very mathy classes like formal methods and models, intro. digital logic, and algorithms. i actually ended up doing quite a bit more than that.
yes i did walk up hill, both way, to class --- i had to cross a valley;-)
if you do anything even vaguely useful / interesting with artificial intelligence, you're going to have to apply those several semesters of calculus, probability, statistics, and linear algebra. computer graphics requires at a minimum vector calculus and linear algebra. if you want to work on the physics side of games, you'll need all the standard tools of the physicist's toolbox (lots of calculus, a really solid understanding of linear algebra, plus the actual physics) in addition to a fair amount of numerical analysis (both for optimization purposes and to transform the continuous world of physics into the discrete world of the computer).
i will admit that these are fairly specialized areas. on the other hand, the problem solving skills should transfer rather directly. after almost picking up a second major in math (i decided to go with a minor instead), i can tell you that each major branch of math has a different flavor and employs different techniques. the mental agility required to apply techniques from several different branches to attack the same problem, i would argue, is the hallmark of a good problem solver. i don't know how else to develop that agility without actually doing the exercises.
my best guess is that it was about 50/50, although the 50% citizens were heavily biased towards the new citizen / first generation citizen group rather than more "established" citizens.
i happen to be one of those people on the non-citizen side (although i already had a green card).
if you think hiring is bad for the commercial sector, try the defense side (where i am now). finding people with useful background / degrees for the stuff i do (robotics / computer vision) is difficult to start with. when you add requirements about citizenship / green-card status, it becomes nearly impossible. (the discussion about whether or not it's efficient to use people with those highly specialized backgrounds to do general software development rather than use the people you have more efficiently as part of software development teams is a different argument, one that i've been on the losing side of at my current group for a long time).
nitpicking, i know, but really what you've described are the virtues of a good software engineer, not so much a good computer scientist.
i see software engineering as an answer to "build the solution" whereas computer science is more about answering "what is the solution". then again, i have a fairly old school "c.s. is a combination of applied applied math and applied discrete math" world view.
i graduated with my first c.s. degree during the peak 2003-2004 and i can tell you that about half the people that i graduated with have since burned out and moved on to new careers. i would estimate that an overwhelming majority of the people that i started out with thought that majoring in c.s. would help them earn lots of money. something like 80% of the people that started in c.s. at the same time i did switched majors because they realized that c.s. wasn't for them. about half the people that were left were people that realized, too late, that c.s. wasn't for them but they were so far down the road that switching majors wasn't an option. most of them ended up having to take the upper division theory classes a few times before barely earning a passing grade, and then got out as fast as they could. they were uniformly miserable.
i stuck around to work on a m.s. in c.s. and i noticed a similar, although less severe pattern there --- again, about half the people that were in my grad foundation sequence classes (compilers, operating systems, algorithms, and a.i.) washed out before they managed to finish the sequence. an informal survey of people in my o/s class showed that about 60% of them were there for the money. just like undergrad, the people who washed out were miserable.
by way of comparison, the people who survived to take the "fun" grad level classes (computer vision, intro robotics, image processing, etc.) were a lot more fun to be with and generally a lot more excited about what was going on. classes went from enrollments of 45-60 to 10-20, professors were markedly more relaxed, and i felt that, in general, i got a lot more out of those classes than i did anything else in my education.
in the long term, i think that c.s., like most of the math / science / engineering disciplines, is extraordinarily demanding and unless it's something that a person really enjoys doing, i don't seem them surviving in a c.s. related career for very long.
i can tell you that while the cars may be "ready" by 2018 (possibly, maybe --- there's some really hairy control theory that needs to be worked out w/ regards to multiple vehicles interacting with each other), there's a whole brave new world of insurance and legal issues that will delay any real adoption for at least another decade after that. the good news is that in the mean time, you're likely to start seeing some very cool vehicle safety technologies start trickling down from the top end of the market (things like automatic cutoff detection, automatic lane tracking, automatic sign recognition, etc.)
there are at least two other companies that have been awarded prime contracts for major system components (pratt & whitney / rocketdyne and atk thiokol). lockheed is supplying the raison d'etre in terms of the orion crew vehicle.
it's impossible to tell if this guy is patenting the idea of doing vision on gpus (in which case there's prior art going back to before 2004 and probably even beyond that in the gpgpu community) or if he's talking about some tremendously clever collection of algorithms that happens to map well to gpu hardware. either way, i suspect that the poor guy is about to discover the hard way just how extraordinarily difficult this problem is.
Maybe open source programmers should spend more time getting their programs to do X instead of just telling people "you shouldn't be doing X at all." --- i think you hit the nail on the head with this. unfortunately, the primary customer in oss is almost always the guy who got the itch to write the program in the first place rather than any representative sample of potential end users.
I've never had problems with documents hundreds of pages in length. I'm calling BS on this as a Slashdot urban legend.
i've had endless troubles with pagination of documents that are *tens* of pages long (25 to be precise). every time i work on a proposal and start putting art in, i lose several hours fiddling around with graphic anchors so that pages have reasonable amounts of text on them. the collective heartburn caused by word during proposals has, in fact, become so severe that there are several of us that are seriously looking into using latex as an alternative.
no, actually, i can't "plainly" see inside the car. i can barely see inside the car, even after having dicked around with my gamma settings. what i can easily see is a ton of reflections of the environment (which are highlighted quite clearly if you do play around with the gamma).
i use infrared cameras on a regular basis for my work and i've never seen near ir behave substantially differently from a monochrome sensor. the near ir cameras that i've worked with (without active illumination) are pretty useless at night. so unless there's some other kind of processing going (image intensification or what have you), i don't see how near ir helps to solve the problem. on the other hand, mid and longwave ir cameras are actually useful so long as you have a reasonable temperature differential between the target and ambient (which is likely to be a problem in places like dc with high humidity and ambient temperatures pretty close to body temperature for long stretches during the summer).
doesn't the wavelength of the lidar depend on the laser that's used? for what it's worth, i've used microbolometer fpa long wave ir (800-1300nm odd) and glass is almost totally reflective. it caused a lot of heartburn (turns out that computer vision in ir bands is a very different problem than in visible spectrum).
... through glass (which is almost totally reflective for the long wave ir cameras that i've used). i wonder if there's something special about the glass they use in vehicles...
Slashdot Mirror
does ireland have a legal concept similar to common carrier in the u.s.? i'm not a lawyer, much less an expert on the irish legal system, but it would seem to me that this case could only work in a country where common carrier laws are either non-existent or very weak. if ireland does have something like common carrier that would cover eircom then a win appears to essentially invalidate common carriers and make any isp that sends traffic through ireland potentially liable, even if both ends of the infringing connection are outside of irish jurisdiction.
you must work for lenovo (http://www.computerworld.com/action/article.do?command=viewArticleBasic&articleId=9034318)
haar, 1910, is the earliest example of a wavelet in a paper. however, you could argue that the core concept (projection onto a functional basis) goes back to fourier, with haar's real contribution being finding a particular functional basis that obeys the rules associated with wavelets. to the best of my knowledge, most of the wavelet math was worked out relatively recently by mallat, daubechies, etc. starting in the '80s
your company hired a loser.
i don't think that one anecdotal data point really means that hiring people with proven records of delivering quality results on time and training them is a bad idea so much as it underscores the importance of not hiring losers. quite frankly, if it takes someone much more than a month to three months to develop a productive level of proficiency in a new language, you're dealing with a dud and it's time to let them go.
i think they were hoping that you'd see the patterns in the exercises and figure it out for yourself.
the tricks that i used most frequently were:
when you see tan(x) by itself, you probably want to rewrite it as -(-sin(x)/cos(x)).
keep a look out for things that look like trig identities particularly when dealing with rational expressions (e.g (tan^2(x)+1)/csc(x) = sec(x)/csc(x) = (1/cos(x))/(1/sin(x)) = cot(x) = cos(x)/sin(x) ==> du/u) -- your goal is to use those identities to bring things into a form that you know how to deal with.
you want usually want to pull things like cos^n(x)sin^(n+1)(x) out by themselves (==> 1/(n+1)u^(n+1)du) by themselves. if you've got unbalanced powers of sin and cos, you probably want to try a power reduction identity to reduce one of them to something a bit more manageable or see if you can do something with the product rule.
your mileage will, of course, vary.
i can't speak for your t.a.s, but my experience was this: i worked my ass off in labs trying to cover the gap between what the lectures had covered and what i thought the students needed to know to do the projects and homeworks. as a result, the people who showed up regularly to labs never scored less than a B on the projects and programming assignments. people who didn't come to lab were all over the map. likewise, i worked my ass off to be prepared for office hours. the students who came to my office hours regularly all passed and did a lot better on average than the people in the class who didn't come. there's no t.a. that can force you to come to office hours or make you ask questions like, "if you were grading a problem like this on the exam, what do you think the professor will be looking for when it gets graded".
for what it's worth, if you're memorizing rather than learning the small handful of tricks that you need to do the trig integrals, you've missed the point.
on the other hand, you can occasionally luck out like i did --- i teach robots how to see and the gov't pays me handsomely for it. i can't really thing of a more awesome job :-) getting those awesome jobs requires a combination of technical mastery that goes well beyond what an o'reilly book would cover, excellent soft skills (communication, interpersonal, leadership), and the ability to self-promote. that last one is, i think, extraordinarily unnatural for most technical people but tends to come more easily to business types.
your point about the problem starting in k-12 education is well taken, and i certainly agree.
i agree that the math core and science can be a major stumbling block when taught poorly --- lord knows i've had more than my fair share of poor math and science profs. that said, there's a major disconnect between how high school in the us fails, miserably, to prepare students for college level math and the fact that math at a university level actually is a lot more than quality time with your ti-89. there are a lot of people who approach university math with a high school math mindset, people who are used to sailing through with no intellectual challenge because teachers are teaching to the test with cookbook answers. my first university math class was a real wake-up call for me in this regard, although i was lucky enough to wake up early enough that i didn't fail spectacularly.
;-)
as a t.a. (so someone who's been on the other side), i had several students come to fight with me when they got their papers back because they felt they deserved more points because their answer and the real answer had some of the same symbols in the same places. it's right, or it's wrong, and while i did what i could to give as much credit as i could for the level of mastery they had demonstrated, there was no way that i was going to give them more credit than they deserved. most of them deserved to fail.
i suspect that this is what you're running into: the combination of people actually expecting you to perform well (i.e. at a far higher level than what high school expected) and no longer being able to slide by without putting effort into the classes.
honestly, i'm surprised you're getting away with so little math and science. we were required to do calc1-3, discrete math 1, linear algebra 1, prob and stats for engineers (what would have been two semesters in the math department crammed into one, with a sadist of a prof), and numerical methods. we also had 3 semester of lab science, including at least 1 sequence (i did chem1-2 and bio1). this doesn't cover the very mathy classes like formal methods and models, intro. digital logic, and algorithms. i actually ended up doing quite a bit more than that.
yes i did walk up hill, both way, to class --- i had to cross a valley
if you do anything even vaguely useful / interesting with artificial intelligence, you're going to have to apply those several semesters of calculus, probability, statistics, and linear algebra. computer graphics requires at a minimum vector calculus and linear algebra. if you want to work on the physics side of games, you'll need all the standard tools of the physicist's toolbox (lots of calculus, a really solid understanding of linear algebra, plus the actual physics) in addition to a fair amount of numerical analysis (both for optimization purposes and to transform the continuous world of physics into the discrete world of the computer).
i will admit that these are fairly specialized areas. on the other hand, the problem solving skills should transfer rather directly. after almost picking up a second major in math (i decided to go with a minor instead), i can tell you that each major branch of math has a different flavor and employs different techniques. the mental agility required to apply techniques from several different branches to attack the same problem, i would argue, is the hallmark of a good problem solver. i don't know how else to develop that agility without actually doing the exercises.
my best guess is that it was about 50/50, although the 50% citizens were heavily biased towards the new citizen / first generation citizen group rather than more "established" citizens.
i happen to be one of those people on the non-citizen side (although i already had a green card).
if you think hiring is bad for the commercial sector, try the defense side (where i am now). finding people with useful background / degrees for the stuff i do (robotics / computer vision) is difficult to start with. when you add requirements about citizenship / green-card status, it becomes nearly impossible. (the discussion about whether or not it's efficient to use people with those highly specialized backgrounds to do general software development rather than use the people you have more efficiently as part of software development teams is a different argument, one that i've been on the losing side of at my current group for a long time).
that's how the c.s. departments in both of the universities that i've attended came about --- they grew out of the math department.
nitpicking, i know, but really what you've described are the virtues of a good software engineer, not so much a good computer scientist.
i see software engineering as an answer to "build the solution" whereas computer science is more about answering "what is the solution". then again, i have a fairly old school "c.s. is a combination of applied applied math and applied discrete math" world view.
i think you're right.
i graduated with my first c.s. degree during the peak 2003-2004 and i can tell you that about half the people that i graduated with have since burned out and moved on to new careers. i would estimate that an overwhelming majority of the people that i started out with thought that majoring in c.s. would help them earn lots of money. something like 80% of the people that started in c.s. at the same time i did switched majors because they realized that c.s. wasn't for them. about half the people that were left were people that realized, too late, that c.s. wasn't for them but they were so far down the road that switching majors wasn't an option. most of them ended up having to take the upper division theory classes a few times before barely earning a passing grade, and then got out as fast as they could. they were uniformly miserable.
i stuck around to work on a m.s. in c.s. and i noticed a similar, although less severe pattern there --- again, about half the people that were in my grad foundation sequence classes (compilers, operating systems, algorithms, and a.i.) washed out before they managed to finish the sequence. an informal survey of people in my o/s class showed that about 60% of them were there for the money. just like undergrad, the people who washed out were miserable.
by way of comparison, the people who survived to take the "fun" grad level classes (computer vision, intro robotics, image processing, etc.) were a lot more fun to be with and generally a lot more excited about what was going on. classes went from enrollments of 45-60 to 10-20, professors were markedly more relaxed, and i felt that, in general, i got a lot more out of those classes than i did anything else in my education.
in the long term, i think that c.s., like most of the math / science / engineering disciplines, is extraordinarily demanding and unless it's something that a person really enjoys doing, i don't seem them surviving in a c.s. related career for very long.
i can tell you that while the cars may be "ready" by 2018 (possibly, maybe --- there's some really hairy control theory that needs to be worked out w/ regards to multiple vehicles interacting with each other), there's a whole brave new world of insurance and legal issues that will delay any real adoption for at least another decade after that. the good news is that in the mean time, you're likely to start seeing some very cool vehicle safety technologies start trickling down from the top end of the market (things like automatic cutoff detection, automatic lane tracking, automatic sign recognition, etc.)
at 500ft altitude?
there are at least two other companies that have been awarded prime contracts for major system components (pratt & whitney / rocketdyne and atk thiokol). lockheed is supplying the raison d'etre in terms of the orion crew vehicle.
if it makes you feel any better, i was cheering for salieri
"C'mon, do you really believe that a four year old Mozart sat down at the piano by himself and composed an opera while drinking earl grey tea?"
No. Mozart was long dead before Earl Grey tea was known as such (see http://en.wikipedia.org/wiki/Earl_Grey_tea and compare with http://en.wikipedia.org/wiki/Mozart)
it's impossible to tell if this guy is patenting the idea of doing vision on gpus (in which case there's prior art going back to before 2004 and probably even beyond that in the gpgpu community) or if he's talking about some tremendously clever collection of algorithms that happens to map well to gpu hardware. either way, i suspect that the poor guy is about to discover the hard way just how extraordinarily difficult this problem is.
there isn't a jury around that wouldn't buy a temporary insanity plea in that case.
not that i have a raging psychotic hatred for dell tech support.
Maybe open source programmers should spend more time getting their programs to do X instead of just telling people "you shouldn't be doing X at all." --- i think you hit the nail on the head with this. unfortunately, the primary customer in oss is almost always the guy who got the itch to write the program in the first place rather than any representative sample of potential end users.
I've never had problems with documents hundreds of pages in length. I'm calling BS on this as a Slashdot urban legend. i've had endless troubles with pagination of documents that are *tens* of pages long (25 to be precise). every time i work on a proposal and start putting art in, i lose several hours fiddling around with graphic anchors so that pages have reasonable amounts of text on them. the collective heartburn caused by word during proposals has, in fact, become so severe that there are several of us that are seriously looking into using latex as an alternative.
probably related to eye safety
no, actually, i can't "plainly" see inside the car. i can barely see inside the car, even after having dicked around with my gamma settings. what i can easily see is a ton of reflections of the environment (which are highlighted quite clearly if you do play around with the gamma).
i use infrared cameras on a regular basis for my work and i've never seen near ir behave substantially differently from a monochrome sensor. the near ir cameras that i've worked with (without active illumination) are pretty useless at night. so unless there's some other kind of processing going (image intensification or what have you), i don't see how near ir helps to solve the problem. on the other hand, mid and longwave ir cameras are actually useful so long as you have a reasonable temperature differential between the target and ambient (which is likely to be a problem in places like dc with high humidity and ambient temperatures pretty close to body temperature for long stretches during the summer).
doesn't the wavelength of the lidar depend on the laser that's used? for what it's worth, i've used microbolometer fpa long wave ir (800-1300nm odd) and glass is almost totally reflective. it caused a lot of heartburn (turns out that computer vision in ir bands is a very different problem than in visible spectrum).
... through glass (which is almost totally reflective for the long wave ir cameras that i've used). i wonder if there's something special about the glass they use in vehicles...