I wonder how much of these improvements in accuracy are due to fundamental advances, vs. the capacity of available hardware to implement larger models and (especially?) the availability of vastly larger and better training sets...
I'm sure all of that helped, but the key ingredient is training mechanisms. Traditionally networks with multiple layers did not train very well, because the standard training mechanism "backpropagates" an error estimate, and it gets very diffuse as at goes backwards. So most of the training happened in the last layer or two.
This changed in 2006 with Hinton's invention of the Restricted Boltzman Machine, and someone else's insight that you can train one layer at a time using auto-associative methods.
"Deep Learning" / "Deep Architectures" has been around since then, so this article doesn't seem like much news. (However, it may be that someone is just now getting the kind of results that they've been expecting for years. Haven't read up on it very much.)
These methods may be giving ANN a third lease on life. Minsky & Papiert almost killed them off with their book on perceptrons in 1969[*], then Support Vector Machines nearly killed them again in the 1990s.
They keep coming back from the grave, presumably because of their phenomenal computational power and function-approximation capabilities.[**]
[*] FWIW, M&P's book shouldn't have done anything, since it was already known that networks of perceptrons don't have the limitations of a single perceptron.
[**] Siegelmann and Sontag put out a couple of papers, in the 1990s I think, showing that (a) you can construct a Turing Machine with an ANN that uses rational numbers for the weights, and (b) using real numbers (real, not floating-point) would give a trans-Turing capability.
I am feeling very grateful that I don't have such a myopic world-view that I believe that everyone participates in the same cultural festivals as I do.-- ~~~~
I just wonder whether the High Priest of the Pastafarians pardoned a pan of pasta for today.
for posts like this that give me tips on how to feel better about myself and my life. And for *anything* that reminds me of all I have to be grateful for, because I find it far too easy to forget.
For the overwhelming majority of us, no matter how bad we think things are, there are others who have it far, far, worse. That's something you can be grateful for even when you're having trouble keeping your chin up.
Although it sounds rather inflammatory and is really, really stupid, the fact that the House has jammed up that committee with people having the intellectual prowess of fleas really doesn't change things.
Are they dull-witted, or merely serving the will of their climate-denying overlords?
this depends on which laboratory you are talking about. most regional toxicology laboratory are able to detect these drugs (usually by LC-MS/MS), and then chemical standards of these drugs are commercially available (despite, perhaps, the big price tag on the chemical standards which are ~1 grand a piece...)
Why does Slashdot have 10x the stories on global warming verses the rapidly increasing US Government debt? Don't you think debt is a more pressing problem?
No.
For some reason the public debt is only an issue when there's a Democrat in the White House. GWB went on a spree of tax cuts and spending, and no one complained that the national debt wasn't being paid down instead.
"Government debt" is a manufactured crisis. There is an attempt afoot to use it as an excuse to bring disaster capital home, so we can experience it ourselves.
3 to 4% of what? Of the total amount of carbon dioxide in the atmosphere right now, about a third is due to humans burning fossil fuels [wikipedia.org]. If we stop emitting carbon dioxide, the warming will level off within a few decades.
I suspect it's too late for that. If we stopped emitting immediately, the world would continue warming due to all the carbon already in the atmosphere, until equilibrium is reached. And that equilibrium might be a good way off, since the continued warming is going to reduce ice cover and thus add more warming, in a vicious cycle, until we reach an equilibrium on that as well.
*Then* we can start waiting for long-term processes to de-carbonize the atmosphere.
Since total human carbon emission is about 3 to 4% (even by IPCC figures)
Percent of what?
So far exactly zero of the 'models' have managed to predict anything
Now you're just talking out your ass...
our 'measures' to combat it seem primarily designed to fill government and large business coffers
...and generating a conspiracy theory to justify it.
And what makes you think it would be much of a challenge to survive? I am amazed by how people seem to confuse inconvenience with survival these days.
Yes, I'm sure you'll be happy to trade your current cushy lifestyle for the dog-eat-dog world of a Mad Max movie.
If you really want to massively cut back carbon emissions, then start rallying against GreenPeace and the other kneejerk 'enviromentalists' blocking of latest generation nuclear power. Rolling out that to replace both old dangerous design reactors and combustion based generation is by FAR the biggest step there could be.
I agree that nuclear power would be a good thing, but the evidence suggests that our civilization isn't mature enough to manage it. Too much profit incentive to cut corners on design, construction, and operations.
1 degree over the next 100 years, 2 degrees over the next fifty years, 4 degrees over the next 25 years. Next year some "scientists" will probably be calling for a 10 degree rise within the next 10 years. Every year, I hear something that sounds less-and-less like hard science coming out of these "scientists" and more-and-more of something that sounds more akin to millennialist religious fanatics proclaiming the end of days.
Hard to see why it's "less-and-less like hard science", since it's based on evidence.
The only invariant in the science of global warming is that it always turns out worse than we expected faster than we expected. If we suddenly find ourselves with an ice-free arctic, we have to take that into account in our projections.
Posting AC because posting anything that even mildly questions GW will get your karma blown into the shitter.
Karma is cheap. You should speak your mind even if it isn't popular.
Slashdot Mirror
I wonder how much of these improvements in accuracy are due to fundamental advances, vs. the capacity of available hardware to implement larger models and (especially?) the availability of vastly larger and better training sets...
I'm sure all of that helped, but the key ingredient is training mechanisms. Traditionally networks with multiple layers did not train very well, because the standard training mechanism "backpropagates" an error estimate, and it gets very diffuse as at goes backwards. So most of the training happened in the last layer or two.
This changed in 2006 with Hinton's invention of the Restricted Boltzman Machine, and someone else's insight that you can train one layer at a time using auto-associative methods.
"Deep Learning" / "Deep Architectures" has been around since then, so this article doesn't seem like much news. (However, it may be that someone is just now getting the kind of results that they've been expecting for years. Haven't read up on it very much.)
These methods may be giving ANN a third lease on life. Minsky & Papiert almost killed them off with their book on perceptrons in 1969[*], then Support Vector Machines nearly killed them again in the 1990s.
They keep coming back from the grave, presumably because of their phenomenal computational power and function-approximation capabilities.[**]
[*] FWIW, M&P's book shouldn't have done anything, since it was already known that networks of perceptrons don't have the limitations of a single perceptron.
[**] Siegelmann and Sontag put out a couple of papers, in the 1990s I think, showing that (a) you can construct a Turing Machine with an ANN that uses rational numbers for the weights, and (b) using real numbers (real, not floating-point) would give a trans-Turing capability.
It's evidence. You expect the FBI to tell them to destroy the evidence?
Read the summary.
Teh FBI wants to have it both ways.
This is the new Corporate Pwned Century. Get used to it.
FTFY
Am I the only one to discover that Firefox doesn't play well with my window manager after the 17.0 upgrade yesterday?
Can't resize any of the windows, and most come up tiny.
Get some real world experience with this.
AC Talks Out His Ass; News At 11
Why do you assume that complaining about a firewall is a liberal bias?
Because if you're conservative, everything that is true or even approaching the truth is "lib'rul bias."
Even the Wall Street Journal.
He's lost himself. What parent meant to say was "NBC didn't report my conspiracy theory about Benghazi.."
I get the impression that for the past month or so FOX has given up on news-spin and just dispenses conspiracy theories now.
I am feeling very grateful that I don't have such a myopic world-view that I believe that everyone participates in the same cultural festivals as I do.-- ~~~~
I just wonder whether the High Priest of the Pastafarians pardoned a pan of pasta for today.
A few nice words can go a long way. Appreciation is important
I try to maintain a policy of: Don't let a day go by without thanking someone who's doing a thankless task.
for posts like this that give me tips on how to feel better about myself and my life. And for *anything* that reminds me of all I have to be grateful for, because I find it far too easy to forget.
For the overwhelming majority of us, no matter how bad we think things are, there are others who have it far, far, worse. That's something you can be grateful for even when you're having trouble keeping your chin up.
A few nice words can go a long way. Appreciation is important
That was a very post, thank you for it.
It very was a post indeed.
or can anyone get in?
You only have to be able to pretend.
In representative democracies, politicians have to please their constituencies, either for real or for pretend. It doesn't matter which.
Although it sounds rather inflammatory and is really, really stupid, the fact that the House has jammed up that committee with people having the intellectual prowess of fleas really doesn't change things.
Are they dull-witted, or merely serving the will of their climate-denying overlords?
This is Slashdot. You're supposed to demand a flying car analogy.
But what does it all *mean*?
Should I buy a Tesla, or give everything I own to the poor and go stand on the mountaintop?
which direction is it pointing?
Slightly to the left, until I see the doctor again.
Oh, great. Now time has a liberal bias too.
We should go back to using cruise missiles and carpet bombing.
Where do you draw the definitional line? Isn't a cruise missile a robot that kills people?
this depends on which laboratory you are talking about. most regional toxicology laboratory are able to detect these drugs (usually by LC-MS/MS), and then chemical standards of these drugs are commercially available (despite, perhaps, the big price tag on the chemical standards which are ~1 grand a piece...)
Thanks.
Why does Slashdot have 10x the stories on global warming verses the rapidly increasing US Government debt? Don't you think debt is a more pressing problem?
No.
For some reason the public debt is only an issue when there's a Democrat in the White House. GWB went on a spree of tax cuts and spending, and no one complained that the national debt wasn't being paid down instead.
"Government debt" is a manufactured crisis. There is an attempt afoot to use it as an excuse to bring disaster capital home, so we can experience it ourselves.
Give in now
I, for one, welcome our new four-degree-warmer overlords!
Temperatures have remained steady for 16 years now
Please look at this: http://en.wikipedia.org/wiki/Instrumental_temperature_record#Warmest_years
and tell us how many of the last 16 years are not included in the 16 hottest years on record.
As the saying goes, you're entitled to your won opinions, but not to your own facts.
in spite of computer models released by the IPCC which show that by now temperatures should be much higher.
And yet for some reason all the world's ice is melting faster than the IPCC ever had the balls to predict.
Why do you need to refer to Gore or bankers, when the scientists are telling you what is happening?
Because it's easier to dismiss the claims of bankers and politicians than the claims of scientists.
3 to 4% of what? Of the total amount of carbon dioxide in the atmosphere right now, about a third is due to humans burning fossil fuels [wikipedia.org]. If we stop emitting carbon dioxide, the warming will level off within a few decades.
I suspect it's too late for that. If we stopped emitting immediately, the world would continue warming due to all the carbon already in the atmosphere, until equilibrium is reached. And that equilibrium might be a good way off, since the continued warming is going to reduce ice cover and thus add more warming, in a vicious cycle, until we reach an equilibrium on that as well.
*Then* we can start waiting for long-term processes to de-carbonize the atmosphere.
Since total human carbon emission is about 3 to 4% (even by IPCC figures)
Percent of what?
So far exactly zero of the 'models' have managed to predict anything
Now you're just talking out your ass...
our 'measures' to combat it seem primarily designed to fill government and large business coffers
...and generating a conspiracy theory to justify it.
And what makes you think it would be much of a challenge to survive? I am amazed by how people seem to confuse
inconvenience with survival these days.
Yes, I'm sure you'll be happy to trade your current cushy lifestyle for the dog-eat-dog world of a Mad Max movie.
If you really want to massively cut back carbon emissions, then start rallying against GreenPeace and the other
kneejerk 'enviromentalists' blocking of latest generation nuclear power. Rolling out that to replace both old
dangerous design reactors and combustion based generation is by FAR the biggest step there could be.
I agree that nuclear power would be a good thing, but the evidence suggests that our civilization isn't mature enough to manage it. Too much profit incentive to cut corners on design, construction, and operations.
1 degree over the next 100 years, 2 degrees over the next fifty years, 4 degrees over the next 25 years. Next year some "scientists" will probably be calling for a 10 degree rise within the next 10 years. Every year, I hear something that sounds less-and-less like hard science coming out of these "scientists" and more-and-more of something that sounds more akin to millennialist religious fanatics proclaiming the end of days.
Hard to see why it's "less-and-less like hard science", since it's based on evidence.
The only invariant in the science of global warming is that it always turns out worse than we expected faster than we expected. If we suddenly find ourselves with an ice-free arctic, we have to take that into account in our projections.
Posting AC because posting anything that even mildly questions GW will get your karma blown into the shitter.
Karma is cheap. You should speak your mind even if it isn't popular.