At Atheist Air, prior to boarding, passengers would be required to spout blasphemous remarks at a display of artifacts from all the major religions. This effectively weeds out anyone who has a secret plan to meet the Creator in the next few hours. Blasphemers would be allowed to carry-on pickaxes, blowtorches, chainsaws, nun chucks, whatever, under the theory that atheists generally try to avoid hurting other people in any situation where there isn't a clear escape route.
being an Engineer means that when you screw up, people die. and
the discipline is no less rigorous than any other kind of engineering. But perhaps its for the best,... Had you been an Engineer, no less than 14 people would have died from that screw up alone.
The (nonlinear) threshold setting on a digital unsharp mask algorithm cause my high pass filter analog to break down, but otherwise it's valid. So ignore the threshold, for a moment, in the unsharp mask. The implementation of the unsharp mask is in the spatial domain as you said, but (without the threshold) it has a dual in the frequency domain. The unsharp mask uses a convolution of the image with a Gaussian for blurring, followed by linear additions and subtractions. Convolution, addition and subtract all have duals in the frequency domain.
The Convolution Theorem states that convolution in one domain (e.g., spatial domain) equals point-wise multiplication in the other domain (e.g., frequency domain). Taking the Fourier transform of a Gaussian function yields another Gaussian function. The frequency domain Gaussian is a low pass filter. The subtraction takes an all pass filter (the original image) and subtracts the low pass filter (the blurred) resulting in a high pass filter. The high pass filter is then added back to the original, making the transfer function of the unsharp mask = 1+a*High PassFilter(f), where a = amount, and the High Pass Filter(f) = 1 - Low Pass Filter(f).
Unsharp masking isn't really the right algorithm to compare deconvolution techniques to. They are apples and oranges in their implementation, strengths and weakness, and mathematical foundations, but like apples and oranges, they are both used for the same purpose. Wikipedia does a fine job comparing them.
Deconvolution, on the other hand, is a direct high pass filter.
The Refocus program I was referring to uses a FIR filter. If by "direct" you mean implemented in the frequency domain then you won't agree that Refocus uses deconvolution.
"A FIR Wiener filter only uses a limited neighborhood of the source pixels and can be easily implemented as a convolution matrix."
FIR Wiener filtering has the following advantages:
Low memory requirements. Only the convolution matrix must be stored.
Ease of implementation. There is no need to do a full Fourier transform.
The transformation is local. The results only depend on a small neighborhood of the original pixel.
The Unsharp Mask and the FIR Wiener filter have all these properties in common.
CSI style surveillance camera enhancement is impossible, but you can get a surprising amount of additional detail out of a blurry photo with properly applied deconvolution. I agree. Check out the deconvolution examples using the Gimp Plug-in Refocus-it which is based on finding the minimum of the error function using Hopfield neural network, or Refocus which is based on a modified form of the Wiener filter, called the FIR (Finite Input Response) Wiener filter. Refocus is conveniently available as a Digikam plugin as well as a gimp plugin.
I've played with Refocus and have had some pretty good results with it, even better than unsharp mask, as the documentation states:
In practice I found that in virtually all cases the results of the FIR Wiener filter were at least as good as those of unsharp masking. The FIR Wiener filter is frequently better in restoring small details.
Unsharp masking is a high pass filter. Blurring, which is a convolution with a Gaussian, is multiplication with a DC centred Gaussian in the frequency domain, a low pass filter. Subtracting that from the original will give you the high frequency components. Unsharp masking adds more of those high frequency components back to the original. The thing about a blurred image is that the high frequencies of interest are by definition of "blurred", not present in the original, so unsharp masking can not restore them.
I've used Refocus for sharpening photographs of text for an OCR, but gocr mostly outputted garbage. I read the article with interest thinking that the fitness function of the particle swarm optimization for blurry scanned text could be the number of English words that are identified when the refocused image is run through an OCR.
Mathworks runs a Matlab programming contest. After the contest is over, they write an analysis of the competition. I was surprised to read what they said about the hacking of one contest:
the first 'hacking' started just after the contest went into daylight, when Hannes Naude was able to clear the beam count via use of a regexp. This started a whole different 'contest' for some of the competitors, who were able to find some very ingenious methods for either returning information from the test suite or crashing the queue. Understanding these attempts can be very educational, and thus I've listed many of them below...
They go on to explain in detail how each of the hacks works. This programming contest encourages learning from others, and the details of these hacks are in the spirit of the open source philosophy. One could learn from code tricks that cheats in this contest to do something else useful.
In Bill Bryson's book A Short History of Nearly Everything, he writes that:
Alexander von Humboldt observed: "There are three stages in a scientific discovery: first, people deny that it is true; then they deny that it is important; finally they credit the wrong person."
It's funny until you realize how true it is. The whole book is extremely interesting, very funny and quite relevant to this topic.
is baffled as to how the it came to be in the country.
Charles Darwin did some experiments on how seeds can be distributed throughout the world. Here's just a sample of his findings:
Drift timber is thrown up on most islands, even on those in the midst of the widest oceans; and the natives of the coral-islands in the
Pacific, procure stones for their tools, solely from the roots of drifted trees, these stones being a
valuable royal tax. I find on examination, that when irregularly shaped stones are embedded in the
roots of trees, small parcels of earth are very frequently enclosed in their interstices and behind
them,--so perfectly that not a particle could be washed away in the longest transport: out of one
small portion of earth thus completely enclosed by wood in an oak about 50 years old, three
dicotyledonous plants germinated.
I can
show that the carcasses of birds, when floating on the sea, sometimes escape being immediately
devoured; and seeds of many kinds in the crops of floating birds long retain their vitality: peas and
vetches, for instance, are killed by even a few days' immersion in sea-water; but some taken out of
the crop of a pigeon, which had floated on artificial salt-water for 30 days, to my surprise nearly all
germinated.
Living birds can hardly fail to be highly effective agents in the transportation of seeds.
Although the beaks and feet of birds are generally quite clean, I can show that earth sometimes
adheres to them: in one instance I removed twenty-two grains of dry argillaceous earth from one
foot of a partridge, and in this earth there was a pebble quite as large as the seed of a vetch. Thus
seeds might occasionally be transported to great distances.
As icebergs are known to be sometimes loaded with earth and stones, and have even carried
brushwood, bones, and the nest of a land-bird, I can hardly doubt that they must occasionally have
transported seeds from one part to another of the arctic and antarctic regions
This paper from MIT describes how to create 3D trees from a series of photographs mostly automatically. It builds what it can from the photographs, and interpolates the rest using L systems.
The last slashdot article on the subject of 3D objects from video sequences does not seem to have a workable approach to creating trees due to their none planar nature.
The Handicap principle explains pretty well why a non-utility gift, such as a diamond, may be prefered over a useful one. It basically says, "Look what an expensive useless item I can buy for you. I am such a good provider that I can afford to squander money." Evolution pressures men to buy crap like this, and pressures women to enjoy crap like this. Evolution created showoffs and their groupies.
It's not my idea though, and as another reply pointed out, it is related to the peacock tail
An example in humans was suggested by Geoffrey Miller who expressed that Veblen goods such as luxury cars and other forms of conspicuous consumption are manifestations of the handicap principle, being used by men to advertise their "fitness" to women.
Miller believes that our minds evolved not as survival machines, but as courtship machines, and proposes that the human mind's most impressive abilities are courtship tools, which evolved to attract and entertain sexual partners.
A downpayment on a house almost fits, but it can't be worn as a status symbol, and really doesn't meet the criteria as a gift.
Voodoo is pretty great, although its automatic feature point estimation in 3D is a bit limited. For features far away, in Free Move mode, the stereo math breaks down and it only gets direction correct, while distance from the camera can be anything from -infinity to +infinity.
An option for creating a textured 3D VRML model from images is PTStereo. It is designed to work with 2 or more images that are taken far apart from each other, not a video sequence. Hugin, and SIFT can create the control points, but PTPicker must be used as the frontend for PTStereo.
The reconstruction of a 3D data set from a series or photographs is pretty cool. Photosynth has this ability as well. The generation of the 3D point cloud is good and all, but I've been looking for a program that goes one step further. Creating a 3D triangle mesh using photographs as textures.
PTStereo does just that. It is part of panotools, but unfortunately the author has not released its source. (PTStereo was only one of a few components of Panotools that is only available in binary form).
PTStereo creates 3D-worlds from sets of photographs. Applications range from object movies to terrain visualization. Any set of images can be used without alignment requirements. 3D-data can be extracted for any feature visible in both images, as long as the images have different viewpoint (non-zero stereo base).
User input involves the triangulation of all images: Corresponding feature points have to be identified, and connected to a mesh consisting of triangles. The output of PTStereo is a 3D-world ready to be viewed with any VRML or 3DMF-browser. This world consists of a texture mapped indexed face set. In addition, world coordinates of all feature points and camera positions are calculated and can be used for measurements.
I've used autopano-sift (with Hugin as a front end) to automatically create and match the features points, and PTPicker's Delaunay triangulation to make the triangles. PTStereo then outputs a VRML file that can be read in to Blender (or any number of VRML reading programs) using this VRML importer.
I now have my 3D triangle mesh textured model in the modeling program. You can do whatever you want with it from there, such as measuring its volume, rigging it to be animated as a CGI element in a movie, analysing the terrain of a mountain to find the best hiking route to a summit, having it be an obstacle for Blender's fluid simulator. There's no limit.
This issue I have with Google Earth, Photosynth and Earthmine, is that you are limited to their datasets, and they tend to only make 3D models of cities and buildings, and low resolution images (compared with photographs) of mountains taken from above only. Sketchup is a step in the right direction, but PTStereo allows you to do Image Based Modeling from your own images.... and it's free!
Better ask a crazy religious fanatic to pilot your top secret SR-71 blackbird than Stringfellow Hawke.
You don't ask a guy with a name like Stringfellow Hawke to fly your top secret black airplane. Why? Because he's obviously going to steal it. He is obviously a prototypical American anti-hero, for fucks sake. He lives in the mountains, he plays the cello, his name is StringFellow Hawke. He can not be trusted. He not going to use the SR-71 to execute American foreign policy. He's going to keep it for himself. He walks out to his backyard to stare at it every night around sunset, the sight of it filling him with such peace and resident satisfaction that he came to believe the perfect haiku had just seven syllables: SR-71 Blackbird.
Thanks for answering. That's really interesting.
It seems geosynchronous satellites are not the best tool for the job here. They could put some in polar orbits, or maybe use a natural satellite.
At 80 degrees north in Eureka, Nunavut, Canada, you would need to point an antenna horizontally to communicate with a geostationary satellite.
There's a photo of an satellite dish antenna pointing horizontally at the south pole. Is communication with that satellite only possible during certain times of the day?
In Firefox 2, when you type something which isn't recognizably a URI into the location bar, it doesn't use "I'm Feeling Lucky", it uses a subtly different Google search mode called "Browse by Name".
It's easy enough to fix: just go to about:config and change the keyword.URL property from its default value,
From a BBC study in 2004 about how and why we find certain things disgusting.
Blue rarely occurs in nature. We have no reason to find it disgusting because it was never associated with disease threats in our evolutionary past.
Green slimy substances make us think of snot. Mucus can harbour dangerous bacteria and viruses and we produce more of it when we are ill.
Yellow is the colour of pus, a substance that is produced from infected wounds. Pus consists of tissue debris, white blood cells and bacteria - many of which may still be alive. Yellow and red suggests the products of a diseased wound - a clear disease threat.
Chances are the monkey knows this too.
But if he has shown that he has no preference before hand, cognitive dissonance is probably just an mechanism that evolved to reduce the danger of being Buridan's ass. The wasting of time pondering between two equally good choices can reduce a individual's utility. Repeating that time wasting activity every time the same choice had to be made would certainly be an disadvantage and natural selection won't let it go unnoticed.
The brain is a neural network. Experience can form rules that could be very difficult to learn in other ways, such as instruction. Consider "The physics of judging a fly ball" in "The Physics of Sports" (page 40 found through google books).
...the angular accelleration of the fielder's line of sight to the ball provides the strongest initial clues to the location of the eventual landing point.
Depth perception doesn't do much good at those distances, but with enough practice, fuzzy logic rules can be trained up and your neural network will run a control system loop to make you catch the fly ball.
Humans choose the base that best suites their particular needs. Programmers use binary because computer hardware only supports two stable states (bistability). Most of us learn base 10 because we can use our ten fingers intuitively for counting.
When problems that come up that require division, it's easiest to work in a base that is a highly composite number, such as 2, 4, 6, 12, 24, 36, 48, 60, 120, 180, 240, 360 etc.
It's a good idea to have the numbers of hours in a day to be highly composite. We can easily divide a day into two, three, four etc. number of equal chunks. Having 10 hours (or even 100 hours) in a day would sucks.
360 was likely choose for the number of degrees in a circle because it's conveniently close to the number of days in a year, while also having 24 divisors.
Are any of the hardwired numbers programmed directly into our brain or are we using them because they end up being the best tool to do the job?
Gimp 2.3 and above now does Lanczos resampling optionally on all its operations.
This resampling method approximates the ideal sinc filter, and provides an excellent balance between ringing and antialiasing. For downsampling, it removes frequencies about Nyquist rate (antialiasing), and for upsampling it correctly interpolates between sample points based on the Whittaker-Shannon interpolation formula. Resampling allows subpixel accuracy so I'm very happy to see they've included the Lanczos resampling kernel.
It's briefly touched upon on this link from the summary under "Improved display when zooming in or out".
I've implemented a 3D Lanczos filter for CT scan data (to eliminate staircasing when reconstructing an isosurface from marching cubes) and it was cool to view the Gimp's code for inspiration.
The grandmasters guiding the discussion boards biased the voters.
"[move 16] sparked loud grumbling on the bulletin board that Krush had "taken over the game". Those who complained were not overstating Krush's influence; her recommendations were selected every single move from the 10th to the 50th.
But perhaps even greater than the effect of [Kasparov's 35th move] on the position was its effect on the psyche of the bulletin board....As the World Team began to panic in a dangerous position, the flames, insults, and petty bickering reached heights...
(Technical loopholes) Microsoft
someone bragged to the bulletin board that he had tricked MSN into letting him vote multiple times,... the ballot-stuffing method he outlined was indeed workable, as several bulletin board members verified on the 59th move
Both 58...Qe4 and 58...Qf5 looked reasonable, but the bulletin board had analyzed the former to a forced loss, so Krush duly recommended the latter. Due to an e-mail glitch, her recommendation and analysis were not received on time by the MSN site, and voting proceeded for some time
several disgruntled members of the bulletin board, knowing the game to be lost, suggested the 59...Qe1?? move which would quickly lead to a Kasparov victory. This was meant as a protest against Microsoft and the whole competition.
I was surprised to learn that With his 62nd move, Kasparov announced a forced mate found by the computer program Deep Junior. So it was more like Kasparov and Deep Junior vs. 4 grandmasters guiding a few amateurs on a bulletin board.
"If the channel is noisy it is not in general possible to reconstruct the original message or the transmitted
signal with certainty by any operation on the received signal E. There are, however, ways of transmitting
the information which are optimal in combating noise.... It is possible to send
information at the rate C (the channel's capacity) through the channel with as small a frequency of errors or equivocation as desired
by proper encoding."
For audio, the channel capacity of cheap cables are many orders of magnitude greater than the information rate of an audio signal. That allows for a great deal of redundancy and a negligible (but not necessarily zero) error rate.
Slashdot Mirror
Scott Adams had this idea in 2003.
The (nonlinear) threshold setting on a digital unsharp mask algorithm cause my high pass filter analog to break down, but otherwise it's valid. So ignore the threshold, for a moment, in the unsharp mask. The implementation of the unsharp mask is in the spatial domain as you said, but (without the threshold) it has a dual in the frequency domain. The unsharp mask uses a convolution of the image with a Gaussian for blurring, followed by linear additions and subtractions. Convolution, addition and subtract all have duals in the frequency domain.
The Convolution Theorem states that convolution in one domain (e.g., spatial domain) equals point-wise multiplication in the other domain (e.g., frequency domain). Taking the Fourier transform of a Gaussian function yields another Gaussian function. The frequency domain Gaussian is a low pass filter. The subtraction takes an all pass filter (the original image) and subtracts the low pass filter (the blurred) resulting in a high pass filter. The high pass filter is then added back to the original, making the transfer function of the unsharp mask = 1+a*High PassFilter(f), where a = amount, and the High Pass Filter(f) = 1 - Low Pass Filter(f).
Unsharp masking isn't really the right algorithm to compare deconvolution techniques to. They are apples and oranges in their implementation, strengths and weakness, and mathematical foundations, but like apples and oranges, they are both used for the same purpose. Wikipedia does a fine job comparing them.Deconvolution, on the other hand, is a direct high pass filter.
The Refocus program I was referring to uses a FIR filter. If by "direct" you mean implemented in the frequency domain then you won't agree that Refocus uses deconvolution.
"A FIR Wiener filter only uses a limited neighborhood of the source pixels and can be easily implemented as a convolution matrix." FIR Wiener filtering has the following advantages:
- Low memory requirements. Only the convolution matrix must be stored.
- Ease of implementation. There is no need to do a full Fourier transform.
- The transformation is local. The results only depend on a small neighborhood of the original pixel.
The Unsharp Mask and the FIR Wiener filter have all these properties in common.I've played with Refocus and have had some pretty good results with it, even better than unsharp mask, as the documentation states: Unsharp masking is a high pass filter. Blurring, which is a convolution with a Gaussian, is multiplication with a DC centred Gaussian in the frequency domain, a low pass filter. Subtracting that from the original will give you the high frequency components. Unsharp masking adds more of those high frequency components back to the original. The thing about a blurred image is that the high frequencies of interest are by definition of "blurred", not present in the original, so unsharp masking can not restore them.
I've used Refocus for sharpening photographs of text for an OCR, but gocr mostly outputted garbage. I read the article with interest thinking that the fitness function of the particle swarm optimization for blurry scanned text could be the number of English words that are identified when the refocused image is run through an OCR.
It's funny until you realize how true it is. The whole book is extremely interesting, very funny and quite relevant to this topic.
This paper from MIT describes how to create 3D trees from a series of photographs mostly automatically. It builds what it can from the photographs, and interpolates the rest using L systems.
The last slashdot article on the subject of 3D objects from video sequences does not seem to have a workable approach to creating trees due to their none planar nature.
The Handicap principle explains pretty well why a non-utility gift, such as a diamond, may be prefered over a useful one. It basically says, "Look what an expensive useless item I can buy for you. I am such a good provider that I can afford to squander money." Evolution pressures men to buy crap like this, and pressures women to enjoy crap like this. Evolution created showoffs and their groupies.
It's not my idea though, and as another reply pointed out, it is related to the peacock tail A downpayment on a house almost fits, but it can't be worn as a status symbol, and really doesn't meet the criteria as a gift.Voodoo is pretty great, although its automatic feature point estimation in 3D is a bit limited. For features far away, in Free Move mode, the stereo math breaks down and it only gets direction correct, while distance from the camera can be anything from -infinity to +infinity.
An option for creating a textured 3D VRML model from images is PTStereo. It is designed to work with 2 or more images that are taken far apart from each other, not a video sequence. Hugin, and SIFT can create the control points, but PTPicker must be used as the frontend for PTStereo.
Blender can import the textured vrml.
PTStereo does just that. It is part of panotools, but unfortunately the author has not released its source. (PTStereo was only one of a few components of Panotools that is only available in binary form).
I've used autopano-sift (with Hugin as a front end) to automatically create and match the features points, and PTPicker's Delaunay triangulation to make the triangles. PTStereo then outputs a VRML file that can be read in to Blender (or any number of VRML reading programs) using this VRML importer.I now have my 3D triangle mesh textured model in the modeling program. You can do whatever you want with it from there, such as measuring its volume, rigging it to be animated as a CGI element in a movie, analysing the terrain of a mountain to find the best hiking route to a summit, having it be an obstacle for Blender's fluid simulator. There's no limit.
This issue I have with Google Earth, Photosynth and Earthmine, is that you are limited to their datasets, and they tend to only make 3D models of cities and buildings, and low resolution images (compared with photographs) of mountains taken from above only. Sketchup is a step in the right direction, but PTStereo allows you to do Image Based Modeling from your own images.... and it's free!
Better ask a crazy religious fanatic to pilot your top secret SR-71 blackbird than Stringfellow Hawke.
You don't ask a guy with a name like Stringfellow Hawke to fly your top secret black airplane. Why? Because he's obviously going to steal it. He is obviously a prototypical American anti-hero, for fucks sake. He lives in the mountains, he plays the cello, his name is StringFellow Hawke. He can not be trusted. He not going to use the SR-71 to execute American foreign policy. He's going to keep it for himself. He walks out to his backyard to stare at it every night around sunset, the sight of it filling him with such peace and resident satisfaction that he came to believe the perfect haiku had just seven syllables: SR-71 Blackbird.
Thanks Ernie Cline.
Thanks for answering. That's really interesting.
It seems geosynchronous satellites are not the best tool for the job here. They could put some in polar orbits, or maybe use a natural satellite.
At 80 degrees north in Eureka, Nunavut, Canada, you would need to point an antenna horizontally to communicate with a geostationary satellite.
There's a photo of an satellite dish antenna pointing horizontally at the south pole. Is communication with that satellite only possible during certain times of the day?
In Firefox 2, when you type something which isn't recognizably a URI into the location bar, it doesn't use "I'm Feeling Lucky", it uses a subtly different Google search mode called "Browse by Name".
It's easy enough to fix: just go to about:config and change the keyword.URL property from its default value,
http://www.google.com/search?ie=UTF-8&oe=UTF-8&sourceid=navclient&gfns=1&q=
to something like
http://www.google.com/search?ie=UTF-8&oe=UTF-8&btnI=&q=
which should restore the "I'm Feeling Lucky" functionality and get you back to normal.
From a BBC study in 2004 about how and why we find certain things disgusting.
Chances are the monkey knows this too.But if he has shown that he has no preference before hand, cognitive dissonance is probably just an mechanism that evolved to reduce the danger of being Buridan's ass. The wasting of time pondering between two equally good choices can reduce a individual's utility. Repeating that time wasting activity every time the same choice had to be made would certainly be an disadvantage and natural selection won't let it go unnoticed.
There's similar research done on how fish catch insects falling out of the air.
It's a good idea to have the numbers of hours in a day to be highly composite. We can easily divide a day into two, three, four etc. number of equal chunks. Having 10 hours (or even 100 hours) in a day would sucks.
360 was likely choose for the number of degrees in a circle because it's conveniently close to the number of days in a year, while also having 24 divisors.
Are any of the hardwired numbers programmed directly into our brain or are we using them because they end up being the best tool to do the job?
Gimp 2.3 and above now does Lanczos resampling optionally on all its operations.
This resampling method approximates the ideal sinc filter, and provides an excellent balance between ringing and antialiasing. For downsampling, it removes frequencies about Nyquist rate (antialiasing), and for upsampling it correctly interpolates between sample points based on the Whittaker-Shannon interpolation formula. Resampling allows subpixel accuracy so I'm very happy to see they've included the Lanczos resampling kernel.
It's briefly touched upon on this link from the summary under "Improved display when zooming in or out".
I've implemented a 3D Lanczos filter for CT scan data (to eliminate staircasing when reconstructing an isosurface from marching cubes) and it was cool to view the Gimp's code for inspiration.
Egos
The grandmasters guiding the discussion boards biased the voters.
- "[move 16] sparked loud grumbling on the bulletin board that Krush had "taken over the game". Those who complained were not overstating Krush's influence; her recommendations were selected every single move from the 10th to the 50th.
- But perhaps even greater than the effect of [Kasparov's 35th move] on the position was its effect on the psyche of the bulletin board....As the World Team began to panic in a dangerous position, the flames, insults, and petty bickering reached heights...
(Technical loopholes) Microsoft- someone bragged to the bulletin board that he had tricked MSN into letting him vote multiple times,
... the ballot-stuffing method he outlined was indeed workable, as several bulletin board members verified on the 59th move
- Both 58...Qe4 and 58...Qf5 looked reasonable, but the bulletin board had analyzed the former to a forced loss, so Krush duly recommended the latter. Due to an e-mail glitch, her recommendation and analysis were not received on time by the MSN site, and voting proceeded for some time
- several disgruntled members of the bulletin board, knowing the game to be lost, suggested the 59...Qe1?? move which would quickly lead to a Kasparov victory. This was meant as a protest against Microsoft and the whole competition.
I was surprised to learn that With his 62nd move, Kasparov announced a forced mate found by the computer program Deep Junior. So it was more like Kasparov and Deep Junior vs. 4 grandmasters guiding a few amateurs on a bulletin board."If the channel is noisy it is not in general possible to reconstruct the original message or the transmitted signal with certainty by any operation on the received signal E. There are, however, ways of transmitting the information which are optimal in combating noise.... It is possible to send information at the rate C (the channel's capacity) through the channel with as small a frequency of errors or equivocation as desired by proper encoding."
For audio, the channel capacity of cheap cables are many orders of magnitude greater than the information rate of an audio signal. That allows for a great deal of redundancy and a negligible (but not necessarily zero) error rate.
I wonder what Ebay plans to do with their acquisition of Stumbleupon from May.
There are various possible explanations why violence has declined over the course of human history given in this TED talk by Steven Pinker.
These people need a reality resolver instead of a spreadsheet replacement.
No individual is more valuable than the community.