Tuesday, June 26, 2007
Update on the Kundu image alignments
I have deleted this post for now. The matter is discussed, without comment, here. Feel free to contact me.
Saturday, June 23, 2007
Anatomy of an alleged academic fraud
I have deleted this post, for now (I may restore it later); however, the images I had posted are still available here.
Wednesday, June 20, 2007
Do the math? Redmond can't
Today's entertaining blog post from HuffPo is this one, by Margaret Heffernan, complaining about how Microsoft hates its users. Apparently the problem is they changed the way equations are handled in the equation editor of Word 2007. So: "Now, if you use send in your articles with equations, they turn, miraculously, into unreadable graphics - so Science and Nature can't print them."
Now, equations have always sucked in Microsoft Word's equation editor; but presumably they looked fine in Word 2007 when the authors submitted them. So the problem is not with Word, but with the conversion software that Science and Nature use. I haven't seen Word 2007's output, but it has to be an improvement on previous efforts: nothing could be worse.
in fact, an experienced user will immediately recognise that the the image of the equation at the top of that post was typeset not in Word, but in TeX. This program was written, single-handedly, by Donald Knuth in the 1970s, revised by him in the 1980s, and has scarcely been touched since then. It is about as bug-free as a program can be; and more to the point, it produces not only beautifully typeset mathematics, but beautifully typeset text -- far superior, in typographic terms, to any but the most expensive commercial typesetting programs. (None of which can come close in the quality of math typesetting.) And it's free, in every sense of the word.
In the form of LaTeX, a macro package that makes TeX easier to use, it has become the standard in the scientific typesetting world; the vast majority of preprints on arxiv.org are typeset with LaTeX.
Here are three examples of equations from the few Microsoft Word-typeset papers on today's cond-mat listing.
From here:
(click on all images for larger versions: the inline images are somewhat randomly resized.) Not particularly awful at first glance, but then you notice the disproportionately tiny integral sign, the inadequate spacing on either side of the "=" sign, and the fraction line that almost bumps into the subsequent "d". And the subscript on the left really ought to be in roman, not italic.
But Word can be much worse than that. Here are examples from from here:
Now, the subscript is quite the wrong size (perhaps you need to adjust the size manually in Equation Editor) and is still in italic, the exponent 2 in the first equation is too large, the integral sign in that equation is ridiculously small and the limits are disproportionately sized, the exponent 2 at the end of the second equation's denominator bumps against the preceding bracket, and (as before) the spacing everywhere is bumpy.
Here are the results of typesetting the same equations in LaTeX (with the default Computer Modern fonts, but one can also use Times Roman or other fonts).
Knuth nailed it a quarter century ago. Microsoft (and other vendors) still can't be bothered to get it right.
Contempt for users? This is what Ms Heffernan should be angry about.
(PS - in case it wasn't obvious, the clipped "L" and brackets in the first equation of the LaTeX output are my fault, not LaTeX's.)
Now, equations have always sucked in Microsoft Word's equation editor; but presumably they looked fine in Word 2007 when the authors submitted them. So the problem is not with Word, but with the conversion software that Science and Nature use. I haven't seen Word 2007's output, but it has to be an improvement on previous efforts: nothing could be worse.
in fact, an experienced user will immediately recognise that the the image of the equation at the top of that post was typeset not in Word, but in TeX. This program was written, single-handedly, by Donald Knuth in the 1970s, revised by him in the 1980s, and has scarcely been touched since then. It is about as bug-free as a program can be; and more to the point, it produces not only beautifully typeset mathematics, but beautifully typeset text -- far superior, in typographic terms, to any but the most expensive commercial typesetting programs. (None of which can come close in the quality of math typesetting.) And it's free, in every sense of the word.
In the form of LaTeX, a macro package that makes TeX easier to use, it has become the standard in the scientific typesetting world; the vast majority of preprints on arxiv.org are typeset with LaTeX.
Here are three examples of equations from the few Microsoft Word-typeset papers on today's cond-mat listing.
From here:
(click on all images for larger versions: the inline images are somewhat randomly resized.) Not particularly awful at first glance, but then you notice the disproportionately tiny integral sign, the inadequate spacing on either side of the "=" sign, and the fraction line that almost bumps into the subsequent "d". And the subscript on the left really ought to be in roman, not italic.
But Word can be much worse than that. Here are examples from from here:
Now, the subscript is quite the wrong size (perhaps you need to adjust the size manually in Equation Editor) and is still in italic, the exponent 2 in the first equation is too large, the integral sign in that equation is ridiculously small and the limits are disproportionately sized, the exponent 2 at the end of the second equation's denominator bumps against the preceding bracket, and (as before) the spacing everywhere is bumpy.
Here are the results of typesetting the same equations in LaTeX (with the default Computer Modern fonts, but one can also use Times Roman or other fonts).
Knuth nailed it a quarter century ago. Microsoft (and other vendors) still can't be bothered to get it right.
Contempt for users? This is what Ms Heffernan should be angry about.
(PS - in case it wasn't obvious, the clipped "L" and brackets in the first equation of the LaTeX output are my fault, not LaTeX's.)
Friday, June 15, 2007
It's ok at Harvard, if you're not Kaavya
What happens when you get caught lifting passages, wholesale, from someone else's book? What if the book you're borrowing from is universally discredited by serious scholars?
If you're Harvard Law School Professor Alan Dershowitz, nothing much happens. You get exonerated by your university, and life goes on.
And you get to successfully lobby against tenure for the man who exposed your plagiarism.
If you're Harvard Law School Professor Alan Dershowitz, nothing much happens. You get exonerated by your university, and life goes on.
And you get to successfully lobby against tenure for the man who exposed your plagiarism.
Saturday, June 09, 2007
Piecing the full picture together
Almost four years ago I picked up a digital camera. I'm not much of a photographer, but I clicked a lot of photos: it costs nothing with a digital camera. To date I haven't printed a single one.
This was a Canon A-80, and served me well till recently, when its colours started going new-age. But before that happened, one of the features that intrigued me was the "panorama mode". You could click multiple overlapping images, and the camera had a "panorama mode" to help you align them (and also to keep the exposure level constant, and so on); and Canon also supplied software to "stitch" these photos together into a larger image.
Unfortunately, the software was Windows-only. I did have a Windows partition on my laptop, but rarely booted into it. But this software didn't even make the pain worthwhile. It provided no fine-tuning capabilities, and the results were rather Dali-esque:
So I abandoned the idea, but still had lots of panorama shots lying around in my laptop.
Recently I tried hugin. It's based on the venerable Panorama Tools, which I found too hard to use when I checked in 2003; but hugin really makes the whole thing much easier. The idea is, you provide "control points" to say where your images overlap (and there's a really neat program called autopano-sift that does this for you, providing output in a format hugin can understand); you optionally indicate lines that should be maintained horizontal (typically, the horizon) or vertical (for example, walls and doors); you choose from various stitching options to correct for various forms of distortion; and, finally, you stitch the program. While hugin can provide the final output by itself, it is recommended that you install enblend, which hugin can then call automatically. Enblend intelligently gets rid of ugly "seams" arising from different light exposure levels, and even deals intelligently with other discrepancies, for example, where people appear in one overlapping frame and not in the other.
It's much slower than Canon's software, but it's quite easy to use after you figure out the basic ideas, and it works beautifully.
Who needs a wide-angle lens?
(The above software claims to work on windows too, but I haven't tried it.)
PS: These examples aren't defect-free. In the first hugin example, the needle is slightly bent (though not as broken as in the Canon stitching). This can be corrected by picking more and better control points. And in the last photo, the same individual appears twice, having been photographed in non-overlapping frames: correcting this is beyond enblend's impressive capabilities.
This was a Canon A-80, and served me well till recently, when its colours started going new-age. But before that happened, one of the features that intrigued me was the "panorama mode". You could click multiple overlapping images, and the camera had a "panorama mode" to help you align them (and also to keep the exposure level constant, and so on); and Canon also supplied software to "stitch" these photos together into a larger image.
Unfortunately, the software was Windows-only. I did have a Windows partition on my laptop, but rarely booted into it. But this software didn't even make the pain worthwhile. It provided no fine-tuning capabilities, and the results were rather Dali-esque:
So I abandoned the idea, but still had lots of panorama shots lying around in my laptop.
Recently I tried hugin. It's based on the venerable Panorama Tools, which I found too hard to use when I checked in 2003; but hugin really makes the whole thing much easier. The idea is, you provide "control points" to say where your images overlap (and there's a really neat program called autopano-sift that does this for you, providing output in a format hugin can understand); you optionally indicate lines that should be maintained horizontal (typically, the horizon) or vertical (for example, walls and doors); you choose from various stitching options to correct for various forms of distortion; and, finally, you stitch the program. While hugin can provide the final output by itself, it is recommended that you install enblend, which hugin can then call automatically. Enblend intelligently gets rid of ugly "seams" arising from different light exposure levels, and even deals intelligently with other discrepancies, for example, where people appear in one overlapping frame and not in the other.
It's much slower than Canon's software, but it's quite easy to use after you figure out the basic ideas, and it works beautifully.
Who needs a wide-angle lens?
(The above software claims to work on windows too, but I haven't tried it.)
PS: These examples aren't defect-free. In the first hugin example, the needle is slightly bent (though not as broken as in the Canon stitching). This can be corrected by picking more and better control points. And in the last photo, the same individual appears twice, having been photographed in non-overlapping frames: correcting this is beyond enblend's impressive capabilities.
Saturday, June 02, 2007
Pilot, armed and dangerous
I have always found the idea of "gated communities", in Bangalore, Delhi, and elsewhere, repugnant. These are mini-townships, large complexes of housing, basic stores and other necessities, into which entry is regulated so that the riff-raff are kept out. They haven't yet made their appearance here in Chennai (though there are advertisements in the paper for "Chennai's first", opening shortly).
Nevertheless, it may be a good idea for certain people. The gates would serve a dual purpose: it would keep these people out of regular society.
Visiting relatives today, we found the street outside their house filled with policemen and crowds of local people. From what we could make out, from the scene and from a report that just showed up on Jaya TV, this is what had happened: a boy (the TV said 15 years old, but he looked much younger) was playing with marbles; it went into the garden (or the gated-community-of-one) of a pilot of Indian Airlines; the boy tried to scale the gate to retrieve it; a dog inside barked; and the pilot emerged from his house and shot the boy.
It seems to have been a non-lethal pellet gun, but sufficient to leave the boy in hospital with a bloody and bandaged chest.
True, the boy was trespassing; and perhaps there were provocations that we aren't aware of. But I'd have thought a pilot, in today's world, would be the last person to resort to guns for an answer, especially in a country where gun crime is thankfully rather rare.
I hope he finds himself in a gated community of a quite different kind.
UPDATE: The "official" story is different. Here is The Hindu's version.
At least one detail in The Hindu is verifiably wrong: the man ("Captain" Elias) is not an "Army Captain" but a civilian pilot. One wonders about the other details, almost every one of which contradicts both the local version and the TV report.
Nevertheless, it may be a good idea for certain people. The gates would serve a dual purpose: it would keep these people out of regular society.
Visiting relatives today, we found the street outside their house filled with policemen and crowds of local people. From what we could make out, from the scene and from a report that just showed up on Jaya TV, this is what had happened: a boy (the TV said 15 years old, but he looked much younger) was playing with marbles; it went into the garden (or the gated-community-of-one) of a pilot of Indian Airlines; the boy tried to scale the gate to retrieve it; a dog inside barked; and the pilot emerged from his house and shot the boy.
It seems to have been a non-lethal pellet gun, but sufficient to leave the boy in hospital with a bloody and bandaged chest.
True, the boy was trespassing; and perhaps there were provocations that we aren't aware of. But I'd have thought a pilot, in today's world, would be the last person to resort to guns for an answer, especially in a country where gun crime is thankfully rather rare.
I hope he finds himself in a gated community of a quite different kind.
UPDATE: The "official" story is different. Here is The Hindu's version.
At least one detail in The Hindu is verifiably wrong: the man ("Captain" Elias) is not an "Army Captain" but a civilian pilot. One wonders about the other details, almost every one of which contradicts both the local version and the TV report.
Friday, June 01, 2007
Subscribe to:
Posts (Atom)