I work with statisticians and modellers, and they have an extremely low opinion of neural nets
Neural networks are very useful to me, but of course, they don't solve all problems and won't bring peace in the world.
When exceptionally good, they will give reasonable interpolations on the data. But forget about correct extrapolations
Neural networks is a synonym for "nonlinear regression". As for any type of regression, the interpolation and extrapolation performance depends on using the right number of parameters. If you try modeling 1000 points with a 500th degree polynomial, don't expect good extrapolation...
When you choose to use 100 dpi fonts, do you think it refers to the dot pitch of your CRT? I don't think so, it means 100 pixels/inch. Besides that what's the point of having 1000 physical dots per inch if you only got 100 pixels per inch.
Well, a 1000 dpi display would be nice, but it would unfortunatly require 100 times more pixels (hence RAM, processing,...) then a current 100 dpi display. That means instead of a 32 Meg video card, your card will need 3 GB of memory (textures will also be 100x larger). On the fastest machines available you'd be getting almost 1 fps in Quake... do I need to continue?
We're not ready for 1000 dpi displays and won't be for at least as much time it'll take to have those displays available commercially.
You could embed an Excel chart in KWord, look at a word document in Konqueror,... What I'm talking about is to make all the Windows work seemingly with Linux apps.
Just a thought... don't know whether that's realistic or not... but couldn't WINE be used (after some heavy modifications, I guess) with XParts to embed Windows apps info Konqueror (or any other app).
GL: Graphics Language. More commonly called OpenGL
Just one precision: GL != OpenGL. What we're talking about here is OpenGL. GL was an SGI-only library, which is now replaced by the more open (obviously) and cross-platform OpenGL. The syntax is similar, but one of the differences is that OpenGL doesn't manage windows by itself, so it can work for X, Windows,...
What's the point of virtual memory when you've got RAM like that
That's the thing: I'm guessing that this machine doesn't use swap at all. There would be so much memory to swap with a (relativly) very small disk bandwidth. For that reason, if you want any performance, you need to put almost everything in memory. The disk is only there when you need permanant storage.
That's true, according to the DMCA, breaking into a computer that has copyrighted software on it is illegal. Therefore, there's no need to fix security holes in windows, since it's illegal to break into a Windows box. No cracker wants to take the risk of being thrown in the same category as those evil people who listen to (their) DVD's using DeCSS, right?
This looks like a move towards having EULA on the security holes themselves: "By agreeing to this EULA, you accept that you will not use any of the security hole in Win 2000 and that you will act as if nothing was wrong..."
I mean, who cares whether the system is secure or not. As long as you agree to the EULA, everyone's safe!
Let us say for the sake of argument that we can all agree on a defined figure, say 1%
That's why I didn't put numbers... OK, now what about the following:
Given the first (relativly unreliable) score, you can calculate the probability of this score being insignificant... if you have 60%, then pick a random number 0x100 and if x60, you draw between the two...
OK, here's what I suggest in order to prevent that from happening again... You have to look at the problem in a scientific/statistical way. When you know that the difference in votes is in the order of 1000 for a whole state, you already know that the result, after recounting, WILL NOT BE STATISTICALLY SIGNIFICANT... whatever you do.
There is then only one solution that makes sense: Put the two names in a hat and draw one. This is as scientific as the recount, but it saves lots of time and money. Of course, since I'm not american, I cannot vote for that law...
It's not even an issue. Those how apply those cracks don't pay for Office anyway. They're after makeing money with buisness, because they need to have valid licenses.
A technology company called "RAMBUS" is claiming to have a patent on DNA. This parent has just been approved by the USPTO.
The company has set its licensing fee to $1 for each DNA string, but offers a quantity rebate and at $10M per individual. The company CEO has been quoted saying "We will protect our IP and will go after any offender". RAMBUS has already asked for a restraining order against 100 million people and will ask the judge to restrain them from using DNA.
When asked the reason for its sudden interest in nuclear technology, a RAMBUS official said: "This is in the line of protecting our IP and making people know we're serious about it. Remember, living free is stealing".
Don't get me wrong... Not only do I disagree with your original comment, but I think it shows that you should learn a bit more about X before you make that kind of comment. I think the best mod would be "+0 Uninformed" (ie. uninformed, but don't change the score or Karma).
OK, why is this moderated -1 Offtopic? I think the guy's dead wrong/uninformed, but it's still his opinion and it's on topic. Moderation is not supposed to be a way of saying you agree/disagree. If it were meant for that, we would have "He's right" and "I disagree"... but I've never seen them in the choices... why is that?
Can anyone explain how the hell it's possible to patent a human gene? I'd like to know that, since you can't fight what you don't understand and I sure don't understand why there isn't just about 1 billion year of prior art (well, ~30000 if it's a human-specific gene, but still)...
whether you can force it down to zero is irrelevent... let's say you can force it down to 1 kbits/s. Text transmissions: no problem. Voice transmission: there's a problem but it can still get through. Video transmission: no way!
The problem is that to remotely pilot an aircraft travelling at Mach 2 (even much below Mach 1), you need a lot of data, and that data needs to be updated rapidly. Think about it. You need a video feed from the cockpit, all the radar info, all the instruments (attitude, altitude, speed,...). There's no way you can get all that through when you're being jammed... Also, if you have 100 planes in the same area, your bandwidth (using spread-spectrum) is divided by 100. That's not much available...
I work with people doing AI research, so I can tell you that we're not yet there. "There" meaning: unmanned autonomous aircraft. That being said, I really believe that in a not so distant future we will be "there"...
Make all your aircrafts unmanned and the next thing you know, your communications are being jammed and all your planes are suddenly down! You can make a communication system more robust to jamming (by using spread-spectrum, and other redunduncy), but you cannot make it totally jamming-proof.
Another problem with such an aircraft is that in order to be guided, the plane needs to emit (so the ground controller can see the data). This has two side effects:
1) The plane cannot be made stealth.
2) It would be all too easy to guide a missile based on those emissions.
I think the future of unmanned aircrafts will depend on AI. Of course, we're not yet there...
Just a correction, there's no problem with a pilot flying an unstable aircraft. Lots of today's aircrafts are naturally unstable, but they are made stable by the control system which takes its input from the pilot. The F-117 stealth fighter is an example (I think it's probably the worse) of an unstable aircraft.
As for the G-force, a computer can take more than a human, but there's still a fundamental limit with what the wings can take, which is probably aroung 15 G's (I'm guessing here), while a pilot can take 7-9 G's.
I've just installed 7.2 and it still has lots of rough edges too. The installer is by far the best I've seen so far, the the software it installs is buggy. I have reported a list of ~10 bugs to the mailing list after the 3rd beta and I've found that very few of them were fixed.
What's probably pissing me off the most is that KLyX, the document processing I use all the time is simply umusable (impossible to open a document!). Now, if they had at least tested it, they would have noticed. Maybe it isn't their job to fix it, but please, don't include a program in a completly useless state.
...which brinds... testing. I've found that the main weakness of Linux is slowly changing. When I first started using it (in, 1995, with a 1.2.3 kernel), the main problem with things not working was missing features (including unsupported hardware). Now that everything has evolved a lot, the problem is with the testing/QA. The released product has all the features, but is full of bugs, which are even more annoying than missing features. This is especially bad for Mandrake 7.1 and 7.2 and that's why, after trying them both, I'm still back with 7.0.
A good complement would be
on
Death March
·
· Score: 2
F. P. Brooks' "The Mythical Man-Month", which gave us Brooks' law: "Adding manpower to a late software project makes it later".
Slashdot Mirror
I work with statisticians and modellers, and they have an extremely low opinion of neural nets
Neural networks are very useful to me, but of course, they don't solve all problems and won't bring peace in the world.
When exceptionally good, they will give reasonable interpolations on the data. But forget about correct extrapolations
Neural networks is a synonym for "nonlinear regression". As for any type of regression, the interpolation and extrapolation performance depends on using the right number of parameters. If you try modeling 1000 points with a 500th degree polynomial, don't expect good extrapolation...
When you choose to use 100 dpi fonts, do you think it refers to the dot pitch of your CRT? I don't think so, it means 100 pixels/inch. Besides that what's the point of having 1000 physical dots per inch if you only got 100 pixels per inch.
Well, a 1000 dpi display would be nice, but it would unfortunatly require 100 times more pixels (hence RAM, processing, ...) then a current 100 dpi display. That means instead of a 32 Meg video card, your card will need 3 GB of memory (textures will also be 100x larger). On the fastest machines available you'd be getting almost 1 fps in Quake... do I need to continue?
We're not ready for 1000 dpi displays and won't be for at least as much time it'll take to have those displays available commercially.
You could embed an Excel chart in KWord, look at a word document in Konqueror, ... What I'm talking about is to make all the Windows work seemingly with Linux apps.
Just a thought... don't know whether that's realistic or not... but couldn't WINE be used (after some heavy modifications, I guess) with XParts to embed Windows apps info Konqueror (or any other app).
GL: Graphics Language. More commonly called OpenGL
...
Just one precision: GL != OpenGL. What we're talking about here is OpenGL. GL was an SGI-only library, which is now replaced by the more open (obviously) and cross-platform OpenGL. The syntax is similar, but one of the differences is that OpenGL doesn't manage windows by itself, so it can work for X, Windows,
What's the point of virtual memory when you've got RAM like that
That's the thing: I'm guessing that this machine doesn't use swap at all. There would be so much memory to swap with a (relativly) very small disk bandwidth. For that reason, if you want any performance, you need to put almost everything in memory. The disk is only there when you need permanant storage.
Oh way, the DMCA is prior art. ;)
That's true, according to the DMCA, breaking into a computer that has copyrighted software on it is illegal. Therefore, there's no need to fix security holes in windows, since it's illegal to break into a Windows box. No cracker wants to take the risk of being thrown in the same category as those evil people who listen to (their) DVD's using DeCSS, right?
This looks like a move towards having EULA on the security holes themselves: "By agreeing to this EULA, you accept that you will not use any of the security hole in Win 2000 and that you will act as if nothing was wrong..."
I mean, who cares whether the system is secure or not. As long as you agree to the EULA, everyone's safe!
Let us say for the sake of argument that we can all agree on a defined figure, say 1%
That's why I didn't put numbers... OK, now what about the following:
Given the first (relativly unreliable) score, you can calculate the probability of this score being insignificant... if you have 60%, then pick a random number 0x100 and if x60, you draw between the two...
OK, here's what I suggest in order to prevent that from happening again... You have to look at the problem in a scientific/statistical way. When you know that the difference in votes is in the order of 1000 for a whole state, you already know that the result, after recounting, WILL NOT BE STATISTICALLY SIGNIFICANT... whatever you do.
There is then only one solution that makes sense: Put the two names in a hat and draw one. This is as scientific as the recount, but it saves lots of time and money. Of course, since I'm not american, I cannot vote for that law...
Seems like we can now transplant just about everything short of the brain
Would you call it "brain transplant" or "body transplant"?
how long will it take someone to crack this?
It's not even an issue. Those how apply those cracks don't pay for Office anyway. They're after makeing money with buisness, because they need to have valid licenses.
but the average slashdotter has better reasoning skills
You mean the "first post" ones or the "naked and petrified" ones? OK nevermind...
A technology company called "RAMBUS" is claiming to have a patent on DNA. This parent has just been approved by the USPTO.
The company has set its licensing fee to $1 for each DNA string, but offers a quantity rebate and at $10M per individual. The company CEO has been quoted saying "We will protect our IP and will go after any offender". RAMBUS has already asked for a restraining order against 100 million people and will ask the judge to restrain them from using DNA.
When asked the reason for its sudden interest in nuclear technology, a RAMBUS official said: "This is in the line of protecting our IP and making people know we're serious about it. Remember, living free is stealing".
Don't get me wrong... Not only do I disagree with your original comment, but I think it shows that you should learn a bit more about X before you make that kind of comment. I think the best mod would be "+0 Uninformed" (ie. uninformed, but don't change the score or Karma).
OK, why is this moderated -1 Offtopic? I think the guy's dead wrong/uninformed, but it's still his opinion and it's on topic. Moderation is not supposed to be a way of saying you agree/disagree. If it were meant for that, we would have "He's right" and "I disagree"... but I've never seen them in the choices... why is that?
Can anyone explain how the hell it's possible to patent a human gene? I'd like to know that, since you can't fight what you don't understand and I sure don't understand why there isn't just about 1 billion year of prior art (well, ~30000 if it's a human-specific gene, but still)...
Is there anyone out there who knows?
But then, you're vulnerable to stealth giant flying scisors!
whether you can force it down to zero is irrelevent... let's say you can force it down to 1 kbits/s. Text transmissions: no problem. Voice transmission: there's a problem but it can still get through. Video transmission: no way!
...). There's no way you can get all that through when you're being jammed... Also, if you have 100 planes in the same area, your bandwidth (using spread-spectrum) is divided by 100. That's not much available...
The problem is that to remotely pilot an aircraft travelling at Mach 2 (even much below Mach 1), you need a lot of data, and that data needs to be updated rapidly. Think about it. You need a video feed from the cockpit, all the radar info, all the instruments (attitude, altitude, speed,
I work with people doing AI research, so I can tell you that we're not yet there. "There" meaning: unmanned autonomous aircraft. That being said, I really believe that in a not so distant future we will be "there"...
Make all your aircrafts unmanned and the next thing you know, your communications are being jammed and all your planes are suddenly down! You can make a communication system more robust to jamming (by using spread-spectrum, and other redunduncy), but you cannot make it totally jamming-proof.
Another problem with such an aircraft is that in order to be guided, the plane needs to emit (so the ground controller can see the data). This has two side effects:
1) The plane cannot be made stealth.
2) It would be all too easy to guide a missile based on those emissions.
I think the future of unmanned aircrafts will depend on AI. Of course, we're not yet there...
Just a correction, there's no problem with a pilot flying an unstable aircraft. Lots of today's aircrafts are naturally unstable, but they are made stable by the control system which takes its input from the pilot. The F-117 stealth fighter is an example (I think it's probably the worse) of an unstable aircraft.
As for the G-force, a computer can take more than a human, but there's still a fundamental limit with what the wings can take, which is probably aroung 15 G's (I'm guessing here), while a pilot can take 7-9 G's.
I've just installed 7.2 and it still has lots of rough edges too. The installer is by far the best I've seen so far, the the software it installs is buggy. I have reported a list of ~10 bugs to the mailing list after the 3rd beta and I've found that very few of them were fixed.
What's probably pissing me off the most is that KLyX, the document processing I use all the time is simply umusable (impossible to open a document!). Now, if they had at least tested it, they would have noticed. Maybe it isn't their job to fix it, but please, don't include a program in a completly useless state.
...which brinds... testing. I've found that the main weakness of Linux is slowly changing. When I first started using it (in, 1995, with a 1.2.3 kernel), the main problem with things not working was missing features (including unsupported hardware). Now that everything has evolved a lot, the problem is with the testing/QA. The released product has all the features, but is full of bugs, which are even more annoying than missing features. This is especially bad for Mandrake 7.1 and 7.2 and that's why, after trying them both, I'm still back with 7.0.
F. P. Brooks' "The Mythical Man-Month", which gave us Brooks' law: "Adding manpower to a late software project makes it later".