Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 


Forgot your password?
Close
typodupeerror
×
Technology

Neural Net Routers To Speed Up Net 85

Posted by CmdrTaco from the who'da-thunk-it dept.
Virek writes "Researchers in the UK have developed a prototype packet router that is potentially much faster than existing digital technology. They are using a combination of a neural network and "diffractive optical components" (er..?). If they are successful in developing a commercial product they predict that this could significantly speed up the net. "
This discussion has been archived. No new comments can be posted.

Neural Net Routers to Speed Up Net More

Neural Net Routers to Speed Up Net

Comments Filter:
  • Alright.. here's my rationale for saying that this cant' do layer 3 lookups.. which is hard to determine conclusively, beause we don't have much information to play with.. luckily we've got some good neural nets upstairs to interpolate with..

    To route data, you have to look at the header of the IP packet and read the destination address and then compare it to the routing table and make a decision of which interface will move the data to the destination the quickest. I don't see any way for an optical device to do this. To read an IP header, you have to translate the light pulses to electrical signals and then pass it to the processor to make the routing decision.

    The way I see this working is it has several streams of date to move from interface to interface, it knows which interface it has to move them to, and this technology is finding the fastest way for the switch to move multiple streams of data across the backplane simulteneously. It would sort of turn the backplane into a totally optical, neural net based ethernet, as it is in a sense a CSMACD (Carrier-Sense Multiple Access Collision Detect), multiaccess media, with the neural net avoiding collisions and finding the best way to transport the data..

    of couse.. my neural net might be returning an invalid result for these inputs..

    //Phizzy
  • you can compress the entire RF spectrum into the color red.. in a manner of speaking. The frequencies that the visible spectrum operates at means there is ALOT more bandwidth available in it. So yes, our eyes are capable of absorbing more information than the most sophisticated antenna... remember.. we're talking nanometers here..
  • That's actually the beauty of the techniques used by companies like Akamai [akamai.com] and Sandpiper [sandpiper.net]. They take the burden off the end user for finding a local mirror. When you try to access a pages on their servers, it automatically selects the closest one to you.
  • About the heat thing :
    AFAIK just about any light-generating device will change a respectable amount of its input energy into heat. When you start using IR lasers, this is even more fun, of course. :)

    Actually, I re-read the article a bit more attentively, and they themselves say that heat generation is a a problem, which I take to mean that it is a *big* problem, since the tone of the article is as optimistic as the average lemming. (though a lot smarter, of course :) )
  • Hmm. (Sorry about repeating you, -brazil-, I just hadn't refreshed before I replied.)

    Our brains use glial cells as insulation. Therefore... it is blasphemous to insulate my wires! I'm monkeying with our lord's highest creation! Woe is me! (Oh, the arrogance.)

    (Wow, thanks for pointing out that technological hubris is, in itself, arrogant. I would never have thought of that myself, ever!)

    -Grendel Drago

  • They are using a combination of a neural network and "diffractive optical components" (er..?).

    Diffractive optics is about controlling light with microstructures. Instead of macroscopic features like lenses, surfaces are modulated in micrometer or even nanometer scale to achieve optical functions that can be impossible to obtain with traditional optics.

    In the case of the router, they probably used diffractive beamsplitters which split one laser beam into say 16x16 equal-intensity beams. They might also use microlenses, which are, like the beamsplitters, basically flat (a few microns high) and thus easy to integrate into an optical system. Mass-producing diffractive optical elements is also cheap and easy, as they usually consist of only one piece of glass or plastic.

    Diffractive optics works best with monochromatic (ie. laser) light, but different kinds of wavelength filters can also be produced (bandpass, add-drop etc.) and are widely used in optical networks (very useful in WDM networks).

    -Janne

  • Hal, can you please find me a route to www.slashdot.org?


    I'm sorry, Dave, but I'm afraid I can't do that...


    AAAGH! Need.... caffiene....

  • I'm afraid you've not kept up on current technology. A laser diode is about the size of a grain of rice, and uses milliwatts of power. Also, laser diodes aren't very expensive (consider that you can buy a laser pointer for $20). While the laser diode in a laser pointer isn't the same as the lasers they are using in this, it's not THAT much different.
  • Sorry, I have to correct you on both points:

    1. Modern outers are in fact generally split into control and forwarding planes. All 'routing' takes place in the silicon of the forwarding plane (just a layer 3 switch), and the job of the control plane is to deal with routing updates etc and populate the forwarding plane with the necessary information as and when that info changes.

    2. The speed of light is 300,000km/sec, so it'll *always* be 40ms RTT between London and New York even if you could fire a laser through a perfect vacuum between them, let alone at the lower speeds in fiber.

    M
  • Nah... IOS is far too illogical to ever be a vulcan
  • don't take me so literal matey

    I was amused by the fact that it's the transition from electrons to photons. taking the genesis story of creation as a metaphor for it.

    Did you really think was expecting to come downstairs one morning to find my 3Com Router making me a cup of coffee!

    jeez, try getting an imagination

    .oO0Oo.
  • Comment removed based on user account deletion
  • The refractive optical component mentionned in the article is an optical grating. It's an array of fine lines that acts much like the surface of a CD. When light hits this, its different wavelenghts are scattered spatially, which is useful in all-optical switching on systems using WDM (Wavelenght Division Multiplexing). Different channels on a same fiber can thus be separated spatially at the switch.
  • What's WRONG with you people?
    It's a semi-joke man!
    ( I'm a spiritual person, myself. Still! good troll, dude)
  • A neural network is useful when (1) you don't know the best explicit algorithmic solution, or (2) implementing the best algorithmic solution would be be too complicated or expensive. So a NN is an approximate solution. It is somewhat of a black box in that you don't know exactly how it is working, nor what its failure boundaries are. However, you can get some idea of how the NN works from its internal weight structure.

  • >This tech looks like it could be separated from a router to
    >be used in an ATM switch, but by the time it is viable,
    >we probably will be using MPLS and POS more than ATM

    It's technology like this that will doom packet-on-wire
    and revive ATM. ATM was made with the principal of
    simplifying the problem before solving it. New
    tricks like this are going to be loads easier to
    implement for ATM with its fixed cell size than
    for packet processing.
  • hmm first up bugger some one else thought of this too. Im a bio/pscyho/computer scientist and i have thought of trying to make one of these before. Lets see if i can answer some of these questions (nd for the first time make it off 1). to the above your absoultly right its probably goign to be a while before we figure out how NN's might work. In between now and then its an art. If we can build one that works the damn might never work again. Or some time or another it might just take a holiday. Diagnosis of these things is going to be close to impossible (maybee we could make a NN to daignose them). I think whats important is to rember that if it works well it works lets be happy at that. Some of these things are quite good at what they do. one of the possible advantages (re above some where) is that we only need one layer the actual packets move through the system and go where they need to go. No examine and switch needed. are NN turing machines, well imho opnion they are rember a turing machine needs infint(or at least unbounded) memory space. With that on in mind you should be able to solve any anolog system using one. How long it takes is just up to you. We are so used to seing T machines as serial computers we dont think there could be some other possibility. Remeber it is possible smash your head against your desk and see!! if this doenst make sense I blame my uni exams
  • In Europe, you can also use propane gas (LPG) to fuel your car. That IS a pressurized gas.

    Difference with Hydrogen is that Hydrogen requires very, very little to start burning (just the presence of Platina is enough) and is has a very very low melting point, somewhere around -290 degrees C. Think of the pressure required to keep that in.

    They're experimenting with all kinds of things, but nothing practical has been found yet.

  • At the bottom you'll find the reference you asked for.

    Just think about this: A neural network performs analog mathematical operations (sums, multiplications and so on.

    A Turing machine is DIGITAL. It can simulate analog computations at a finite level of presition. But something is lost. Remember that some numbers have infinite decimals (1/3=0.333333...).

    But those are only comments to begin to understand. Check the reference!

    ----------

    @TechReport{DasGupta99, author = "Bhaskar DasGupta [slashdot.org] and Georg Schnitger [slashdot.org]", title = "On the Computational Power of Analog Neural Networks", institution = "DIMACS", number = "99-49", month = oct # " 8", year = "1999", url = "f tp://dimacs.rutgers.edu/pub/dimacs/TechnicalReport s/TechReports/1999/99-49.ps.gz [rutgers.edu]", abstract = "We survey the computational power of analog neural networks.", note = "Wed, 27 Oct 1999 21:00:00 GMT", }
  • Big Bang theory for children?! Pop science mags and texts have pretty much ensured that if you're old enough to read, you've heard of the Big Bang theory.

    Besides, you haven't been paying attention... remember the /. story about the robots that tested themselves and evolved in a computer simulation, then synthesized the models they wanted?

    Let's see... testing a design, modifying it slightly and figuring which one tests better, and repeating as necessary, is against God? Boy, does *he* ever suck!

    Seriously, though... I can't believe I'm actually replying to this troll...

    -Grendel Drago

  • I fixed the invalid form key problem. Use lynx!! Good old lynx, nice to know in this day and age there is still a web browser that can work on even neural networks. Ouch, that was a stretch to tie it into the topic.

  • Sounds |\|337, but it's not a router. (Score:5)

    by Phizzy ( 56929 ) on Wednesday June 07, 2000 @04:10AM (#1020267)
    Routers have two components : a routing component and a switching component. The switching component takes input on an interface, moves it across the backplane to the routing component, and then and pushes it out another interface when the routing component is done with it. The routing component reads the destination from the IP header, compares it to it's routing table and decides whether it is routable, and if so, which interface it will go out. It seems to me that this technology would only take care of the level 2 (switching) layer of networking, which is not the major area of concern in networking right now.. With existing electrcal technology, we have backplanes that can push 160Gb/s of IP traffic. That, on one device, is enough to push 16 OC-192s. The problem is if a single device had to push 16 OC-192s at wire-rate (full speed, basically), the route processor would not be able to handle the IP lookups to feed the backplane. So this technology (if they can get it to work.. which it seems like they should be able to) _will_ be useful once we get a route processor that can keep up with it.

    This tech looks like it could be separated from a router to be used in an ATM switch, but by the time it is viable, we probably will be using MPLS and POS more than ATM, but there will always be a need for fast layer 2 technologies.

    I like the current trend in networking to move away from electrical processes in switching.. once we can get a fully fibre-based system we will be able to remove much of the inherent latency in telecommunications.. like the 100ms lag to go across the atlantic, or the 60ms lag to go across the United States. Once we have comparatively few electrical devices taking care of the routing and all optical devices doing the switching, the only thing that will cause lag will be processing and the speed of light.

    //Phizzy
  • Well then someone please call Paramount and find out how they're doing it, so we can all go home early.
  • You can kick your fuel pump. Gasoline is not a pressurized gas -- would you want to drive around with one of those soda-fizz CO2 canisters under your car? Ka-pow! Now imagine the liability issues... brr.

    -Grendel Drago

  • Not bad. Arrogance, obscenity, are you as stupid IRL?

    AC, let me introduce you to something called a "mirror"...

    The neural technique is simply a means of choosing the optimum non-blocking configuration. Go and read a textbook on comms theory and come back when you understand why an M x N non-blocking switch is a hard problem.

    I'm not going to read a textbook to understand an article such as this. And that's the key flaw of the article: I do have solid background knowledge in computers and networks, but I'm not a specialist in that particular field. Now, it may be correct that neural networks are a good thing in that context, but the article totally failed to explain anything.

    Seriously, I'm a CS major, and I did attend the odd lecture about neural networks. From what I've heard there, neural networks are a tool to implement rules the exact details of which you can't specify very well, i.e. image recognition. I fail to see, on a very basic level, why they should be better than a good, solid priority queue for traffic scheduling.

  • What "noise" is there in a big load of network packets wanting to be delivered?

    I'd guess there's things floating around like packets for unavailable servers, duplicate ACKS, broken packets from buggy routers and corrupted packets. TCP handles this quite well from the point of view of the server/client relationship, but not so well for optimal routing.
  • There's a big difference between the switching that this can do and solving the IP lookup problem. The neural net structure in the router (as described) can just route in a grid from one edge cell to another. In most routers, there are relatively few of these "edge cells" -- one per line card, or at most one per port on the various line cards. So say between 16 and 1024 endpoints in their system (and 1024 is bigger than almost all deployed systems). Compare this to the 3 billion plus valid IPv4 addresses, or to the 2^126 IPv6 addresses in assigned spaces, and the structure they're using cannot physically scale to the necessary sizes.

    (My background? I work at a company that designs chips to power the line cards in routers.)
  • 1. Yes, the route processor has to deal with routing updates and various bookkeeping functions mostly, but to route traffic the router still has to read the dest ip and compare it to the routing table (which is generally done on a processor on the interface cards nowadays, not the RP, but is still an electrical function). If you think you can do this optically, let me know.. I would love to see the details of it.

    2. Yeah.. that's basically what I was saying, we will still have to deal with the speed of light as a limitation, but we won't have to deal with the 8ns/bit electrical latency you incurr every time you hit an electrical repeater, ATM switch, etcetcetc..

    //Phizzy
  • You, sir, are a dik-dik.

    And BTW, boloney is not a word.

    You're welcome.
  • Complain about neural networks the next time you use some metal alloy, eh? More than a few have been "discovered" through neural network applications. And let's not think about the untidy world of nuclear wasted disposal. Or anything that does non-linear shortest path finding... Plenty of real-world uses in robotics for AI, among other things... Perhaps you should read some technical articles on Neural Network applications. I'd look at the various IEEE-CS publications. And, of course, that's just Neural Networks. There's all sorts of fun and applied stuff done with Expert Systems, Case-Based Reasoning, Genetic Algorithms&Programming, and more recently, there seems to be a glut of agent-based "stuff".

    Er, and you believed pundits talking about events far in the future? I'd be pretty pissed I didn't fly to work and eat little food-pellets if I thought that way :)

    And After all that... it's unfortunate that either words are taken out of context, or fools get publicized, but that's the way it is in a lot of fields...

  • Dogma to you maybe, but I think it should be pretty self-evident to anyone that has not been brainwashed by the atheists into believing that God is a lie.

    Nope, it's only self-evident for those who have been brainwashed by christian fundamentalists that the bible is literal truth, along with a load of nonsense that said fundamentalists basically just made up.

    The only lie is that of atheism, a lie which convinces you there is no God and so it is alright to lie, cheat, murder and rape since there is no punishment after death.

    Really ethical behaviour should not come from fear of punishment, but from tolerance and friendship towards one's fellow humans.

    This is why crime rates have been so much worse since our once great country spurned the Lord's path.

    Care to explain how the crime rates of some explicitily non-christian countries are actually lower?

    Including the rule about worshipping no other God? If you aren't Christian you are failing that straight away.

    Nope. Atheists don't fail that.

    To be saved, you have to believe in and worship the Lord, which counts out false religions and cults like Islam, Catholicism and Buddihism straight away.

    Just for the record: what freak cult do you belong to? (Please don't bother pointing out that your belief system is not a cult but the Real Truth)

  • What you point out, A good programmer is marked by his or her ability to think about a problem and understand a good (not necessarily the best, since there are too many metrics involved to optimize for all of them) way to solve it. , Case-based reasoning (admittedly an infant field, more or less) attempts to solve this problem. Especially CBR's like Meta-AQUA that reason ABOUT reasoning, in an attempt to understand WHY certain problems exist and HOW they were solved.

    By remembering how a problem was solved, a previously successfull methodology can be applied to a new problem. Problems (of course) arise depending on how you approach the application of previous solutions. A simple transformation may not work (transforming a sort routine from C to Lisp is not very viable, the domains are too different), so instead a derivational approach can be used. This derivation follows the reasoning that was used behind writing, say, a sort routing in one particular language, and not the specific actions.

    This reasoning about reasoning gets you a bit ahead of the neural networks... Not that they don't do a great job, and there is a vaste amount of research just dying to be done on/with them, it's just that they don't currently scale up too well. We don't know a good architecture that begins with a neuron, can create outputs, but can also re-configure itself to deal with new problems. Since a traditional feed-forward network is trained to minimize a particular error function, it doesn't work at all when you introduce a problem that has a completely different error function (Let alone different amounts of input).

    Cognitive Psychology has given us some interesting models of how humans might think, and they're relatively easy to mimic with CBRs (I'm sure I've got researchers laughing on the floor with that one :) They give us a good means of generically solving problems. They are, I think, a step in the right direction, especially in that they show a capablitiy for reasoning on multiple domains, a weakness of neural nets (An instance of a neural net is limited to solving only one specifc problem).

    Ah well, just some crack-addled rambling, eh?

  • Researchers in the UK have developed a prototype packet router that is potentially much faster than existing digital technology. They are using a combination of a neural network and "diffractive optical components"

    Of course the bloody thing is fast --- it uses optics instead of old fashioned electrons. If I were to build an idiot-simple optic router that uses a routing scheme from the 1970s, I'd be willing to bet it would be pretty fast too!

    Stories like this remind me of one of my favorite papers from the Annals of Improbable Research. The Paper is titled "Advances in Artificial Intellegence since the 1970s" (or something like that). it has three German co-authors, and consists of a blank page:)

  • So we got fast routers and nice pipes for the people who can afford them. The end user (us) is probably NEVER going to see the speed increase as most of us are on dial-up. Instead of making new routers and new laser technologies for people who can afford them come up with a new low priced connection for the end user that will allow him to take advantage of the already abundant bandwidth.

    Cable's nice, DSL is too (where it's available) but with people like UUNET installing oc-192's I think we need a new home internet connection that will allow us to get much better speeds.
  • I'm sorry, but that is incorrect. Artificial Neural Networks are not Von Neumann Model-based computing devices. In the first place, they are ANALOG. I has been recently prooved that Artificial Neural Networks have more computational power than Turing Machines (who in turn have the same computational power as Von Neumann Computers). Computational power is the ability to solve classes of problems. There are problems that can be solved with ANNs and at the same time cannot be solved by TMachines. So, Church's thesis was proven to be incorrect!. [s]
  • Making the internet backbone fabric better helps everybody. If you're on a 56k connect and you're hitting a site with 1 million other 56k users, all of you are going to be slower than your max because the pipes probably won't handle the load.

    Plus, it's better to make the backbones bigger than needed now so as more and more people do get broadband home connections it doesn't kill the net.

    But you are correct, we need people to work on home broadband connections in addition to all the backbone work.
  • Now I'm really sorry, because my reply was incorrect, this is the 2.0 version: --- I'm sorry, but that is incorrect. Artificial Neural Networks are Turing Machines. In the first place, they are ANALOG. I has been recently prooved that Artificial Neural Networks have more computational power than Turing Machines Computational power is the ability to solve classes of problems. There are problems that can be solved with ANNs and at the same time cannot be solved by TMachines. So, Church's thesis was proven to be incorrect!. [s]
  • Because, that brain would try to take over and destroy the world. Like in the 50's classic horror flick The Brain That Wouldn't Die...
  • ...but I would prefer having smarter users. Ones that have heard of the concept of "local mirrors", for example.
  • Because of objections from the brain's previous owner
  • For around a year now, I've wondered if there's any advantage to mimicking the visual cortex in networking. To me, it seems like the bandwidth would be immense. However, the range is significantly short with this sort of connection.

    Sci-Fi always contained stories of people with cybernetic parts fused to them. How about the other way around? It would seem we're a more efficient computer than the turing model.
  • Why don't they just put a brain in a jar wired to a T-1 and get it over with?
  • ... At least the "neural" part. The article did not give a satisfying explanation what good it actually does!

    The use of neural network techniques allows us to tolerate considerable amounts of system noise -- indeed, the system requires noise -- and therefore provides a more scalable solution than conventional means.

    Huh?? If noise is a problem, then shield the damn thing! I mean, scheduling and switching aren't exactly about dealing with "dirty" sensorical data, so what the fuck is this guy talking about?

  • it's quite symbolic that we should move from the dark days of electric to the light days of optical.

    Maybe the bible is a book of science fiction and not just a bunch of fairy stories.

    routing via neural networks on a global optical scale might be the final step to sentience that the network needs.

    If the major providers upgrade simultaneously they could create the world in six days and then take a rest day to download porn on their spanking new network.

    This forbidden fruit would taint the network and cast it down into eternal damnation but provide the net with free will.


    .oO0Oo.

  • Going Mainstream (Score:1)

    by Anonymous Coward
    The base technology is pretty cool, although it has a long way to go for industrial srength useage. I hope the BT and Heriot-Watt University enlist some outside help to continue development. BT is historically very slow when it comes to developing and deploying new technology and/or products. What seems to have worked for them in the past with efforts like this is essentiall outsoucing or partnering with a bigname shop to do the majority of the nuts and bolts work, then they slap on that goofy BT logo on it and use their branding to send it off with a good piece of already established marketshare.

    If this hits as a mainstream switching technology, it will allow for smaller OSS's to compete with the big boys like Worldcom, Qwest, Global Crossing and AT&T. Switching seems to be the value component now, not just huge bandwidth availability.

    Posting AC due to that Damn Invalid Form Key Error! Anyone wknow why this is happening?-ttm

  • In general, I like the idea of using light to solve the interconnection problem in neural networks. But...

    Potential is the key.

    Potentially, hydrogen is great fuel for cars and planes, as it stores lots of energy and doesn't have "bad" byproducts. But the minor inconvenience is that we must find a way to transport the stuff and carry it around without becoming walking, driving or flying bombs.

    Same here. They're talking about a few major problems to overcome first. So a commercial version will not be around for years.
  • Me too. I can especially do without the 90% of my company that clogs the network with all sorts of random pictures and chain emails and shit.
    Image all the current technology without your average dumb home user....
    unfortunately it's the money that's being squeezed out of their pockets that bought all the equipment. we should still keep them off tho.
  • Amazing Router
    Please use your brain like powers
    DSL faster

    This morning brought to you by crack

  • Routing/Internetworking (Score:3)

    by Anonymous Coward on Wednesday June 07, 2000 @03:57AM (#1020294)

    This is a great idea. It's hight time the internet moved forward with technologies like this, but the real question is this: are we ending up with too many new technologies, and as a result the demand for qualified individuals - individuals that understand the fundamentals of routing and internetworking and can learn these new technologies and integrate them with current technology, becoming too much - ie, the demand for them outpaces the rate at which individuals that come along with this knowledge become available? I think so, and, working in the industry, I've seen it happen all too often: the Internet in general is facing a problem: As a Slashdotter pointed out in the comments section of the story about the great Internet pioneer, Donald Davies, there are fewer and fewer people who understand Internet fundamentals fully. Yes, there are a few - great hackers, CCIE's, and old-timers. But new IT professionals walk in with their CS degree or MCSE and really don't know much. They probably couldn't even tell you what a /30 or a /29 is. More CCIE's are the solution, some people would say, but the fact is that the CCIE qualification, while good, ties people down to one vendor - Cisco, which will create a kind of Microsoftian situation, except with networking equipment and protocols (ever hear of the Cisco Properietry protocols? Ayup.)

    However, a solution to this problem is at hand. The GNU/Zebra. This is a robust routing daemon which is fully open sourced, released under the excellent GNU license. I can't remember the URL offhand, but there is a large project which aims to implement a CCIE-like course based on the product, and which also includes a lot of GNU/Linux material. The course will take the candidate through Linux basics, and then move on to Networking basics, and then advanced Networking (BGP, IGRP, OSPF, etc). They will then be taught how to apply these concepts with the GNU/Zebra. The reasoning behind this is that long after Cisco has died out, Zebra and Linux will live on, due to their open source nature. It's a common argument against all propreirty companies, and , I think, a valid one. Even if the company is huge, it might not survive in all sectors, and if it is a hegu one concetrating on one sector - it might go down completely with advances in technology and demand for flexible solutions that the company, no matter how big it was or how it tried to cover al l the bases, so to speak, failed to see and got left behind. This isn't unlikely to happen in the Networking arena, and Cisco is a prime target for the kind of failure described above. The ZIE (Zebra Internetworking Expert) course will also be amied at educating more people, even though it will be fairly expensive at first, it will be priced reasonably once it reaches a certain level of acceptance by the business community. Their will be 3 exams. 2 written, and one lab. The 2 written will consist of Linux basics and Networking, (which will include both Networking basics, in the Linux module, intermediate Networking and advanced Networking in the Networking module). In the lab, the candidates will be asked to simulate a real working eviroment and interface with other hardware and software, using the GNU/Zebra technology.

    Availability: These exams will be available through various LUGs around the world, eventually. At first they will probably only be available through a few centers in major areas in the US and Europe.

    Pricing: There are several factors in the price here, and although it will be more expensive overall than the CCIE at first, the prices will go down as the qualification gains acceptance by the business community and the availability goes up with the rise of Linux, Open Source, and cooperating from various LUGs and other Linux/Open Source organizations. The written exams will be around $250 each, with the lab exam costing $1,000. If the candidates fail their first time, they will be given a compensation period of 2 months in which they can rewrite the lab exam for $500. The most expensive part of the course will be travelling to Zimbabwe. It is neccessary for all candidates who have completed the lab and written exams to be initiated in Zimbabwe, where they will need to have sexual intercourse with a Zebra mare in heat. At present, female candidates aren't provided for. This will change as the conditions above are fulfilled , though. The trip itself will cost $5,000, hotel accomodation included.

    It's the hope of the organizers of this qualification that it will encourage the use of non-propreitry solutions for networking in mainstream organizations, and promote the use of Open Source technology in general. It's believed by them that the Open Source methodology will not only lead to technical benefits, but will allow communications to go to the next level by bringing back the cooperation of old that started the internet and allowed it to grow, in short: the hacker and scientest and military cooperation of the 60s, 70s and 80s., without which such innovation would have been impossible

  • Despite the truly staggering quantity of flamebait on this article so far, I think this is really a huge leap forward: eventually, as the limitations of procedural computing and programming become more and more obvious (as if Windows 2000 wasn't enough evidence :-) neural computing will be programming, in the same way that computing used to "be" vacuum tubes and is now silicon.

    Programming will become working out the solution to a problem in a general sense, and then training a neural network to behave in that way. No more if-else statements, very little hard-coding at all (at least, nothing that will get turned into a real "program" as we understand it now. Computing is just getting to big and too fast for an unaided human brain to satisfactorily handle all eventual possibilities and / or inputs, and neural networks excel at this.

    I'm just waiting for the first Artificial Neural Networking Operating System: I figure it'll come around as soon as somebody comes up with a less embarrassing acronym than ANNOS :-)

  • As for what is blasphemous about neural networks, any attempt to make a mockery out of the Lord's works by base imitation is blasphemous

    Ah, so in your opinion, Airplanes are blasphemous for they work by imitating birds' wings?

  • But the systems seems to take care of both of those problems (as I understand them): it takes the signal and simultaneously works out where it needs to go and how it gets there. Like all neural computing, the fact that it does this is non-obvious, but it certainly seems to handle both eventualities.

    If I'm wrong, somebody please enlighten me...

  • I'm with you.. I wish I could see MUCH more info on this.. and thusly I can't answer your first two questions because I want the answer myself.. but the third question is easy.. this won't be hot at all.. it's all optical.. it would require VERY little power because it doesn't seem to be doing any computational work.

    Speculating on the first question, though, I don't think it would have to be that big.. I'm imagining MANY 2d metricies of optical channels on a board/chip/whatever, all of which would be very small and thin (you can put them very close together because they wont interfere with each other), with some given amount of channels per interface on the switch to deal with however many concurrent IP streams are going through the device. As long as the switching itself and the switch decision-making process is quick, which it seems like it would be since it is a simple optical interference function, you wouldn't need a huge amount of channels/interface.

    All speculation... maybe we need an Ask Slashdot with these guys. I would love that.. I'm sure their NDAs wouldn't. though.

    //Phizzy
  • Fear of new technology. Fear of people different from you. (Actually, it's fear that they might be the same as you.) Fear of change. Fear of anything you don't understand.

    I suppose fear of your god must come pretty easily with all the fear.

    You have a book that tells you how to live instead of figuring it out on your own, and I'm the brainwashed one?

    Base imitation? Base imitation?! Well, that cuts out airplanes (imitating birds), clothing (fur), electric lights (the sun), not to mention the fricking wheel (feet).

    If you're such a technophobe, why post to slashdot?

    -Grendel Drago

  • Your description of neural nets is somewhat flawed;

    No, it was merely rather compressed, focusing on the end result.

    the basic idea is that each neuron receives a stimulus, modifies it according to a well-defined but simple set of rules, and passes it on to further neurons. It gets interesting when you allow the neuron to change its internal rules passed on the data (adaptive neural nets).

    Are we talking about stuff like back propagation? isn't that terribly slow?

    This isn't strictly what is happening in this case. The inputs to the neurons are light beams, and their effect is to influence the likelihood of that neuron being chosen as a step of the path through the net. So, the system settles out with one path (hopefully the optimum) being chosen.

    Nothing new to me in all that...

    One reason why it's better than traditional methods is that it's a lot faster (there's no computation at each node of the path, because it relies on a physical effect).

    AFAIK most adaption methods are real speed killers.

    Another problem I see is that you can't make any sort of guarantees that the result will ne what you desire. What if your adaptive neural network adapts itself into a particularly bad position? Neural networks have been known to produce very unexpected results now and then.

  • Dammit, I swear that bold stuff didn't happen in the preview...
  • I made it onto the anti-spammers list with one of my accounts. Do I get a prize?

    --
  • Are you sharing whatever it is you're smoking? I think I'd like some....
  • Sometimes it's funny to see non-programmers comment on the progress of computer programming.

    Usually it's not.

    There's a lot more to programming than just "solving problems" (in the traditional sense, like cracking RC5 or playing chess). For the average application, much more code is devoted to the human interface than is devoted to solving problems. User interfaces cannot be represented solely by neural networks. There are many, many discrete problems where neural networks are suboptimal -- for example, parsing a protocol or performing arithmetic. In these domains, digital logic excels.

    In fact, "neural networks" as you describe them are very little more than genies in your computer. You have to train the nets to solve problems -- but first, you need to define the problem, figure out the inputs and outputs, and decide on a reasonable set of training data. You overlook all of this, and represent ANNs as being able to solve anything you can represent in English. The truth is far from that. Have you ever tried to tell a person how to perform a complicated task? How many times did it take you until they understood it? How much longer until they did things the way you thought it should be done? No matter how good your computer's neural network is, it will be either too smart, too dumb, or too specialized to solve a lot of problems given only vague descriptions (at least for the next 20 years or so).

    A good programmer is marked by his or her ability to think about a problem and understand a good (not necessarily the best, since there are too many metrics involved to optimize for all of them) way to solve it. This solution may involve fuzzy AI like ANNs. It may involve traditional imperative/procedural programming. It may involve functional programming. There are too many different tasks that computers take on to say there is one best methodology.
  • Adaption as speed killer? Hm... no, on-line training of a neural network can be quite fast. In what is implemented in the router, there would be no need at all for adaption.

    This is not a traditional feed-forward network trained through back-propagation of error. This is a competitive neural network along the lines of a Kohonen, Grossberg, or Hamming network. What you've been reasoning about is a feed-forward network, which is a completely different beast altogether.

    My suggestion: Actually read a book on the subject matter. An excellent one is "Neural Network Design" by Hagan, Demuth, and Beale. It's quite handy, and does an excellent job explaining the how's and the why's...

  • Just one more step in the Internet's quest to become a sentient entity and destroy all the meatheads. I can't wait. Oh... wait...
  • "My router is having a brainfart!"
  • Oh, I see. You subscribe to your ideals -- "Don't Imitate The Divine Plan!" -- which sound really nice when shouted loudly, and ignore all the nasty little side-effects, like the wheel being evil, and Against God.

    I know, I know, it's a troll. But it's fun!@

    -Grendel Drago
  • I hate to mention this, but the article does not say, or even imply, that it will be used for IP transmission. It merely said telecommunications. I believe this system will be working on telephone routing, not internet traffic. Big difference. Different protocols, different routing paterns, and more consistent traffic. My understanding is that it is using lazers to optically calculate which trunks are to have how much throughput to which other trunks. Basically a hardware implementation of a software algorithm.

    Thats just my .02

  • Or...

    When the NN can make a quicker guess than an exponentially complex (expensive, time consuming) algorithm.

    This is cool in situations where a good guess is all you need and you don't need axact answers

  • Me too. I can especially do without the 90% of my company that clogs the network with all sorts of random pictures and chain emails and shit.

    Perhaps if you explained to your boss how non-business-related network usage is getting in the way of your online pornography usage, some sort of fairer policy could be implemented?

  • It's an interesting concept, but I'm missing a lot of information.
    • How much space does the bloody thing take up ? with a laser for each neuron, I can imagine things getting pretty huge
    • Isn't this thing going to be hidously expensive ?
    • Can you fry eggs on this or what ?
  • Wasn't there a story about some Russian guy who had his brain hooked up to run a telephone exchange?

    (Of course, I use 'had it hooked up' in the involuntary sense...)

    -Grendel Drago

  • I see a few problems immediately with the scheme devised here... The prolog.INS modulator is not bound closely to the main threading abstractor, and the concept of using a KDI interface to a dsucked intulator is laughable at best.

    Oh yeah, Mozilla is a slow buggy piece of shit anyway, so it's irrelevant how fast these new routers are.

  • If you aren't a Christian, you cannot be ethical. This is simple logic - since God has laid down ethical rules for us to follow through the Bible, anyone who is not a Christian does not follow these rules and cannot help but be immoral and unethical and indeed destined for Hell.

    That's not logic, that's dogma, and stupidly interpreted dogma, too. Now even if we assume for a moment that the bible lays down the only valid set of ethical rules (which it does not, as it's inconsistent and self-contradicting as hell), there's quite obviously nothing that prevents someone to follow these rules without being a christian...

  • No, the problem is that the solution of most real-world problems requires the use of all that head cruft we accumulate by living in the real world. To have a machine solve these problems, they must either be reduced to an algorithm (like what's been done with chess) or emulated by emulating the human hardware, which no one's figured out how to do, which is why babelfish is so fscking hilarious.


    -Grendel Drago

  • This can't be happening (Score:3)

    by Wellspring ( 111524 ) on Wednesday June 07, 2000 @04:26AM (#1020317)

    I am sure this is some kind of hoax. Just ask Minsky. Neural networks don't and can't do anything. They are a dead end, and research must be supressed as Perceptron research was before them. You foolish mortals have no idea what a dangerous^H^H^H^H^H^H^H^H^H silly idea neural networks are. After all, if it trains itself, how can I credit my genius for its success?

    (Humor)

  • Ones that have heard of the concept of "local mirrors", for example.

    On the contrary, this would presuppose that users chose their own routers, which is simply not the case. Perhaps there should be greater caching at ISPs, but this is difficult in a world of personalization, streaming content, and pages dynamically generated from real time data feeds.

    Technologies such as those offered by Akamai [akamai.com] claim to be able to address this. And, IIRC, it runs on Linux.

  • I fail to see, on a very basic level, why they should be better than a good, solid priority queue for traffic scheduling.

    Possibly because you're never going to be able to implement a priority queue which will be effective for all kinds of traffic w/o having some kind of dynamic adaptation to the changing traffic characteristics?

  • San Jose - To a crowd of tech reporters and industry heavyweights Sisqo Systems unveiled its newest thong router - the NTNG2000. NTNG2000 is designed as a gateway for the millions of thong packets flying through cyberspace. Thong Song crooner and CEO Sisqo said, "Baby, this is the hottest router in the world. It's routing the vida loca."

    "Thong packets are the hottest segment of the market," said industry analyst Bob Smeckins. "Sisqo once again proves to be the industry leader."

    Neutral thong packet usage has been growing at a rate of 34% a month for the last year since its development by R & B stars Dru Hill. A competing format by Microsoft - G-STRING has been announced but has yet to be seen. "We don't see G-STRING as a threat to neutral thong packets, and we are working closely with Microsoft to develop G-STRING routers also. If they decide not to cooperate then we will kick there bu-butts," said Sisqo while busting some phat moves on the stage.

    The NTNG2000 is housed in a black box with silver metallic trim. The target market appears to be the yet untapped inner city youth demographic.

  • Oh frick, diagnostic nightmare (Score:4)

    by Morgaine ( 4316 ) on Wednesday June 07, 2000 @10:37AM (#1020321)
    One thing that distinguishes neural networks from other approaches is that the "cleverness" (ie. the effective rulebase that determines what happens) is embedded implicitly in the internal state of the network as a set of trigger weightings, and the pattern of weightings has absolutely *ZERO* obvious relationship with the work being performed in any sense that is meaningful to humans.

    It's bad enough trying to diagnose routing problems right now, in a world of dynamic but still fairly deterministic routing algorithms. Add neural networks into the equation and all bets are off: we'll never be able to determine why any particular route was taken. (An explanation of "because those 57 nodes fired" is not particularly helpful).

    Oh joy.

  • 300c (Score:1)

    by crm0922 ( 50203 )
    blast those lasers through a cesium tube and BANG packets from the future....

    Chris
  • "Telephone routing" is increasingly a historical term. Today, IP is frequently carried over ATM backbones along with voice data. This adds some burstiness to the traffic patterns. For traditional voice connections, the number of endpoints is still hugely more than their scheme could scale to. It might be managable routing based only upon the ATM Virtual Path, but unfortunately, not all routes use that. Sometimes the ATM Virtual Circuit must come into the equation, and that multiples the 256 (or 4096, depending on whether you're on the edge of the ATM network or not) VPs by 65536 VCs. Maybe you could restrict the domain sufficiently to solve it using their scheme, but they'd have to work at it.

    My two cents: it's spelled "laser," not "lazer."
  • The most expensive part of the course will be travelling to Zimbabwe. It is neccessary for all candidates who have completed the lab and written exams to be initiated in Zimbabwe, where they will need to have sexual intercourse with a Zebra mare in heat.

    Does anyone even read comments before moderating them anymore?

  • Sorry boyz
    Neurons not only respond to an *ill defined* set of rules, but they *like* them! They adapt on their own according to whatever synaptic traffic is thrown at them. There was an article on this in a VERY early dr dobbs ('76? '77?) , which did a beautiful job of describing the comm model neurons use in nature, and an analog model for same.
    The big problem with true neural nets is 'seeding' them with the right initial conditions so that they converge on what you want them to do. It's a chaotic computation problem when it's pure, and that is a very hairy area. 8P
    All about attractors and stuff... You have to 'train' them.
    Do I make sense here?
    If not; Bugger off!
  • Bwahahaha! Excellent!
    (High fives brilliant troll!)
    Rock on!
  • Uh... What's so much more dangerous about Hydrogen in comparison with gasoline??
  • I don't care how big a lego castle you build, it's not going to become sentient. Routers aren't smart. Even these are extremely special-purpose. It doesn't even get input from the world. (As far as I can tell, it doesn *not* read packets aside from the headers. Neural networks can do dull, repetitive tasks, too.

    Wishful thinking is nice, but this one's just not meant to be.

    -Grendel Drago

  • Neural networks are adaptive systems. They can "learn" to tolerate noise and other problems by reprogramming themselves.

    What frigging' noise, for christ's sake?? What "noise" is there in a big load of network packets wanting to be delivered?

Slashdot Top Deals

The moon is made of green cheese. -- John Heywood

Close