I haven't read TFA, but I was just reading the actual scientific article before I came here. Your objection doesn't quite apply for a similar reason to what you mention: the software was separately trained for each of the nine subjects. For each subject you give them a set of different words (it was 60 words, five from each of 12 different categories of concrete nouns) and record an fMRI of each one.
Now what you do is drop two of those out, train the software on 58 words, and see how well it can guess which words the other 2 were. They show that they are pretty good at predicting the two untrained words.
The interesting bit is how they do this: they basically form a mathematical space with 25 dimensions corresponding to simple sensory-motor features (e.g. eating, nearness, touching, pushing, running, etc.). It is then a matter of using linear regression to break down the brain's responses to each of the trained words into this 25 dimensional space. Then you can turn the system around the other way and use semantic information about the words (calculated based on covariance of work occurrence form a large corpus of text) to predict what the brain activity should be for a given word, then pick the word whose predicted pattern is closest to the experimentally determined pattern.
It is pretty neat work.
Most interesting is that they tried a number of random 25-element bases instead of the hand selected one and showed they offered much lower predictive ability. That supports theories that cortical representation of the words does in fact break down into sensory-motor categories of the type they hand selected.
Yes. People forget that experiments can produce results showing an effect that seems real even if the effect is not real, just through dumb luck, basically. Then only those successes are reported and all of a sudden it looks like there is a real effect where there is none. Whenever this comes up I direct people to the file drawer problem.
Suspicious binary systems might also be a sign of bigger problems, like in Greg Egan's Diaspora (one of the coolest novels I've read in a long time). PANIC!
No, you're making the mistaken assumption that advertisement is inherently bad, and, in particular, that advertisements cannot be entertaining. But neither of those is true, e.g. giving away free samples of food is advertising and is one form most people don't mind.
You can't just claim that trailers do not count as content to be anticipated, because people clearly do view some trailers this way (the OP, for instance). Advertising is just some thing that spreads awareness of a product. As such it is only good or bad in particular instances.
For some reason the my comment where I actually explain what's going on is not modded up to where people will see it. A memristor M satisfies the equation dphi = Mdq. It is one of the six differential equations that can be written that way taking two of the four fundamental circuit variables current, voltage, charge, and magnetix flux. The other five were well described and embodied in electrical elements: this sixth one finishes the symmetry.
Figure 1 in the paper explains it. The four fundamental circuit variables are current, voltage, charge, and magnetic flux. There are six ways of choosing two of these four, which correspond to differential equations relating the variables. Two of them are "given" in that charge is the time integral of current and magnetic flux the time integral of voltage: dq = idt. dphi = vdt.
As for the others, they are components. For instance, a resistor R fits in dv = Rdi. A capacitor C fits in as dq = Cdv. An inductor as dphi = Ldi, and a memristor fills in the missing dphi = Mdq.
Speaking of Yahoo! answers, this (the last one) is the funniest incorrect answer I've seen so far:
Question:
"What is the meaningof "corrolary" in this sentence?
------------------
As a result, oil demand becomes less and less responsive to movements
in international crude oil prices. The *corrolary* of this is that
prices would fluctuate more than in the past in response to future
short-term shifts in demand and supply."
Answer:
"Comparable to corollary in a heart, central blood vessel. Could say
"heart of the matter" or point.
Wikipedia has this very useful system whereby one can actually provide a "reference" for statements (it is so useful that some scientific journals are considering requiring that submitters include these "references," too).
My point is that it is not a problem with Wikipedia; it is a problem with accepting unsourced statements.
But there is also the issue that Wikipedia is often right and we humans tend to approximate the probability that a source provides correct information and trust any new information from that source per that probability. A side effect is that if one fails to continue to verify the verity of a fact, one continues to use an old estimate of the trustworthiness of the source. But this is not Wikipedia's fault.
As others have pointed out, more or less: shouldn't the school be teaching the students instead of complaining that the students don't know important things?
I distinctly remember learning about proper sources way back in elementary school, and then in middle school, and in high school.
I more or less agree with those statements, but you've just swapped your use of the word random away from random access memory (which is what we were talking about and which doesn't mean items in memory must be retrieved at random) to talk of random processes.
Clearly elements in an associative memory can be retrieved at random by using noise as a retrieval key and that is more akin to the type of randomness you've just mentioned (and with which I tend to agree).
I take it you mean the brain, but my instinct is to say that brain memory should be considered content addressable (i.e. associative) rather than random-access, but perhaps they are not strictly opposites.
Certainly, though, I can't think of any random access processes in the brain. On the other hand, sequential access is a reasonable claim, at least for procedural (performance of learned sequences of actions) and episodic memory (memory for personal events in a spatiotemporal context, implying the ability to recall "the next moment in time" given a starting point).
""If you have an application written for any multi-core or single processor architecture that's written to work with Linux, you can take it, compile it and have it running on our chip in minutes," he said. "Now, if you want to ratchet up the performance, we provide libraries and interface mechanisms that customers can use to tune code.""
from here
The depressed patients did just as well as the healthy controls on a spatial working memory task, plus they actually showed up to the test in the first place (as I mentioned in a different comment), which both suggest that the patients were actuall motivated to perform the task.
The research paper this press release is based on also makes clear that all the subjects had been off medication for 4 weeks (give or take, I don't recall exactly).
Slashdotters seem to have very little respect for the research and peer-review system that papers have to go through to be published. As if the questions that occur to people in the first 30 seconds on hearing some subject don't occur to the people spending months doing the research...
1. Correct, no experienced FPS players: "Given a likely relationship between familiarity with video games and the outcome measure, individuals reporting high expertise in video games were excluded"
I am probably not free to copy the whole paragraph about the program, but here's the gist: The program was a virtual reality town. On day 1 subjects got 20 minutes of orientation then 30 minutes navigating around the town to destinations selected by the computer. Their ability to find specific locations was then tested, and if they didn't perform well they got 30 more minutes of practice. Three days later they got 20 minutes to get used to the program again. Two to four hours later their memory of locations in the town was tested (they were tasked to navigate to a new set of locations, different from the specific destinations used on day 1).
Keep in mind this is not a game (despite./'s title), it is just a virtual navigation task. They don't say, but I expect it was just using arrow keys to move around the virtual town. Not much learning curve.
2. They didn't have to memorize many routes. The whole virtual city (from the figure they show) is basically a big, curved X shape with maybe 2-3 other side roads in total. They just had to learn the basic set up of the town so they could go back to a location when asked to.
Regarding the "less mental power", the depressed subjects performed just as well as healthy controls on a spatial working memory task. The distinction is important: the game task tests navigation memory learned over more or less 2-3 days (plus the short refresher on the day of testing), working memory involves manipulating things online. If anything, the latter is probably more challenging (I could be wrong, though, I haven't done the two tasks myself).
3. I dunno, 30 minutes of testing doesn't seem like long enough to really reduce motivation. They must be somewhat motivated in the first place, though, to even show up for the two days of testing.
depression as in "I've been depressed for the past 4 months..."
it takes more than a day for hippocampal volume to reduce, and thus to reduce spatial memory.. also, it is very likely the two have different underlying mechanisms (occasional sadness vs. clinical depression)
I admit I only gave the paper a quick read, so I can't say for sure. But my impression was that spatial information was only discarded in passing information to the next layer in the model. That strikes me as reasonable. For one, they're simulating the dorsal stream, which, in my understanding, is basically attended-object specific, so it seems proper to discard the relationship between the attended object and the rest of the scene. As for discarding spatial relationships between two features of the same object, that also strikes me as roughly reasonable. In real brains there isn't a strict tree-like hierarchy, projections from one region go both to the next higher region but also skip past it and go to yet higher regions. Thus if we have projections A->B->C, B can discard the spatial relationship of two units in A, as long as A also projects to C, which would then still get the spatial information from A as well as the combined information from B (hope that makes sense). It's true that they didn't include such connections in this model, though. I still think it's fair, at least as a starting point for more complex models.
They do discuss the lack of feedback projections, but I also think it's fair to ignore those for the present purposes, because feedback makes things a lot more complicated, modeling-wise.
Finally, I don't have time to go back and check this, but it seemed like the SVM was used to classify the output of the network. That is, it struck me as a test to see how well the highest layer in the network ended up representing the input (after all, you need *some* way to see how well it's doing, and that's a straightforward way). Could be wrong, though.
Well, I may have to get a bit speculative here. First, procedural memory is independent of semantic and episodic memory (look up the famous patient HM who lost his episodic memory from surgery for epilepsy but was still able to form new procedural memories).
Procedural memory develops more slowly than episodic/spatial memory (indeed, by its very nature episodic memories only take one exposure to form). This was shown by Packard and White (1996) who let rats perform a task that could be solved with either procedural or spatial memory and then lesioned either the hippocampus or some part of the basal ganglia (important for procedural memory) at different times to see when the rats switched from one strategy to the other.
Of rhymes I'm not sure, as I'm not up to date on language things. Other mnemonic devices in which one associates items to be remembered with other items (like the method of loci from ancient Greece) are more episodic or spatial memory based in that they rely on forming one-shot associations between, e.g., spatial locations and items to be remembered.
The time course for consolidation from episodic to semantic memory is still disputed: sometimes it seems to take very little time (days), but sometimes, especially for HM who was mentioned above, it may seem to take years, but in all cases it appears to be sort of a gradient rather than a sudden jump from episodic to semantic. I think it's safe to say it takes at least a few days, and in particular, it seems to happen during sleep (unfortunately I don't remember if it's slow-wave sleep or REM sleep).
Slashdot Mirror
Texas?
You may be interested in the term subitization: the ability to rapidly determine numerosity for small sets of items.
I haven't read TFA, but I was just reading the actual scientific article before I came here. Your objection doesn't quite apply for a similar reason to what you mention: the software was separately trained for each of the nine subjects. For each subject you give them a set of different words (it was 60 words, five from each of 12 different categories of concrete nouns) and record an fMRI of each one.
Now what you do is drop two of those out, train the software on 58 words, and see how well it can guess which words the other 2 were. They show that they are pretty good at predicting the two untrained words.
The interesting bit is how they do this: they basically form a mathematical space with 25 dimensions corresponding to simple sensory-motor features (e.g. eating, nearness, touching, pushing, running, etc.). It is then a matter of using linear regression to break down the brain's responses to each of the trained words into this 25 dimensional space. Then you can turn the system around the other way and use semantic information about the words (calculated based on covariance of work occurrence form a large corpus of text) to predict what the brain activity should be for a given word, then pick the word whose predicted pattern is closest to the experimentally determined pattern.
It is pretty neat work.
Most interesting is that they tried a number of random 25-element bases instead of the hand selected one and showed they offered much lower predictive ability. That supports theories that cortical representation of the words does in fact break down into sensory-motor categories of the type they hand selected.
But I bet that TMS has a similar effect on computers...
Drunk people can't possible eat *that* much.
It is this very attachment and craving that keeps you from attaining it.
Yes. People forget that experiments can produce results showing an effect that seems real even if the effect is not real, just through dumb luck, basically. Then only those successes are reported and all of a sudden it looks like there is a real effect where there is none. Whenever this comes up I direct people to the file drawer problem.
Suspicious binary systems might also be a sign of bigger problems, like in Greg Egan's Diaspora (one of the coolest novels I've read in a long time). PANIC!
You can't just claim that trailers do not count as content to be anticipated, because people clearly do view some trailers this way (the OP, for instance). Advertising is just some thing that spreads awareness of a product. As such it is only good or bad in particular instances.
For some reason the my comment where I actually explain what's going on is not modded up to where people will see it. A memristor M satisfies the equation dphi = Mdq. It is one of the six differential equations that can be written that way taking two of the four fundamental circuit variables current, voltage, charge, and magnetix flux. The other five were well described and embodied in electrical elements: this sixth one finishes the symmetry.
As for the others, they are components. For instance, a resistor R fits in dv = Rdi. A capacitor C fits in as dq = Cdv. An inductor as dphi = Ldi, and a memristor fills in the missing dphi = Mdq.
Question:
"What is the meaningof "corrolary" in this sentence?
------------------
As a result, oil demand becomes less and less responsive to movements in international crude oil prices. The *corrolary* of this is that prices would fluctuate more than in the past in response to future short-term shifts in demand and supply."
Answer:
"Comparable to corollary in a heart, central blood vessel. Could say "heart of the matter" or point.
The (point) of this is that prices..."
My point is that it is not a problem with Wikipedia; it is a problem with accepting unsourced statements.
But there is also the issue that Wikipedia is often right and we humans tend to approximate the probability that a source provides correct information and trust any new information from that source per that probability. A side effect is that if one fails to continue to verify the verity of a fact, one continues to use an old estimate of the trustworthiness of the source. But this is not Wikipedia's fault.
As others have pointed out, more or less: shouldn't the school be teaching the students instead of complaining that the students don't know important things?
I distinctly remember learning about proper sources way back in elementary school, and then in middle school, and in high school.
Clearly elements in an associative memory can be retrieved at random by using noise as a retrieval key and that is more akin to the type of randomness you've just mentioned (and with which I tend to agree).
You're in more trouble than you thought.
I take it you mean the brain, but my instinct is to say that brain memory should be considered content addressable (i.e. associative) rather than random-access, but perhaps they are not strictly opposites. Certainly, though, I can't think of any random access processes in the brain. On the other hand, sequential access is a reasonable claim, at least for procedural (performance of learned sequences of actions) and episodic memory (memory for personal events in a spatiotemporal context, implying the ability to recall "the next moment in time" given a starting point).
FWIW:
""If you have an application written for any multi-core or single processor architecture that's written to work with Linux, you can take it, compile it and have it running on our chip in minutes," he said. "Now, if you want to ratchet up the performance, we provide libraries and interface mechanisms that customers can use to tune code."" from here
as others have said, yeah: it runs linux
they also have a compiler for it (of course), an IDE, a "full-system simulation model", etc.. lots of development tools..
not that I've seen any of them, but they claim to have a nice development environment..
Only $435 for 10,000 units. Are there 9,999 people on here who want to go in on that?
The depressed patients did just as well as the healthy controls on a spatial working memory task, plus they actually showed up to the test in the first place (as I mentioned in a different comment), which both suggest that the patients were actuall motivated to perform the task.
The research paper this press release is based on also makes clear that all the subjects had been off medication for 4 weeks (give or take, I don't recall exactly).
Slashdotters seem to have very little respect for the research and peer-review system that papers have to go through to be published. As if the questions that occur to people in the first 30 seconds on hearing some subject don't occur to the people spending months doing the research...
Based on the paper at http://ajp.psychiatryonline.org/cgi/content/full/1 64/3/516 I can sort of address these points..
./'s title), it is just a virtual navigation task. They don't say, but I expect it was just using arrow keys to move around the virtual town. Not much learning curve.
(all quotes are from the paper)
1. Correct, no experienced FPS players: "Given a likely relationship between familiarity with video games and the outcome measure, individuals reporting high expertise in video games were excluded"
I am probably not free to copy the whole paragraph about the program, but here's the gist: The program was a virtual reality town. On day 1 subjects got 20 minutes of orientation then 30 minutes navigating around the town to destinations selected by the computer. Their ability to find specific locations was then tested, and if they didn't perform well they got 30 more minutes of practice.
Three days later they got 20 minutes to get used to the program again. Two to four hours later their memory of locations in the town was tested (they were tasked to navigate to a new set of locations, different from the specific destinations used on day 1).
Keep in mind this is not a game (despite
2. They didn't have to memorize many routes. The whole virtual city (from the figure they show) is basically a big, curved X shape with maybe 2-3 other side roads in total. They just had to learn the basic set up of the town so they could go back to a location when asked to.
Regarding the "less mental power", the depressed subjects performed just as well as healthy controls on a spatial working memory task. The distinction is important: the game task tests navigation memory learned over more or less 2-3 days (plus the short refresher on the day of testing), working memory involves manipulating things online. If anything, the latter is probably more challenging (I could be wrong, though, I haven't done the two tasks myself).
3. I dunno, 30 minutes of testing doesn't seem like long enough to really reduce motivation. They must be somewhat motivated in the first place, though, to even show up for the two days of testing.
not depression as in "aw, I feel sad today"
depression as in "I've been depressed for the past 4 months..."
it takes more than a day for hippocampal volume to reduce, and thus to reduce spatial memory.. also, it is very likely the two have different underlying mechanisms (occasional sadness vs. clinical depression)
They do discuss the lack of feedback projections, but I also think it's fair to ignore those for the present purposes, because feedback makes things a lot more complicated, modeling-wise.
Finally, I don't have time to go back and check this, but it seemed like the SVM was used to classify the output of the network. That is, it struck me as a test to see how well the highest layer in the network ended up representing the input (after all, you need *some* way to see how well it's doing, and that's a straightforward way). Could be wrong, though.
The title of the magazine is Science. The magazine announced its breakthrough of the year. So: Science's Breakthrough of the Year.
Well, I may have to get a bit speculative here. First, procedural memory is independent of semantic and episodic memory (look up the famous patient HM who lost his episodic memory from surgery for epilepsy but was still able to form new procedural memories).
Procedural memory develops more slowly than episodic/spatial memory (indeed, by its very nature episodic memories only take one exposure to form). This was shown by Packard and White (1996) who let rats perform a task that could be solved with either procedural or spatial memory and then lesioned either the hippocampus or some part of the basal ganglia (important for procedural memory) at different times to see when the rats switched from one strategy to the other.
Of rhymes I'm not sure, as I'm not up to date on language things. Other mnemonic devices in which one associates items to be remembered with other items (like the method of loci from ancient Greece) are more episodic or spatial memory based in that they rely on forming one-shot associations between, e.g., spatial locations and items to be remembered.
The time course for consolidation from episodic to semantic memory is still disputed: sometimes it seems to take very little time (days), but sometimes, especially for HM who was mentioned above, it may seem to take years, but in all cases it appears to be sort of a gradient rather than a sudden jump from episodic to semantic. I think it's safe to say it takes at least a few days, and in particular, it seems to happen during sleep (unfortunately I don't remember if it's slow-wave sleep or REM sleep).
Hope that helps!