Slashdot Mirror

What then is the reaction to the lifting force? Remember all forces require an equal and opposite reaction - Newtons laws still apply at this scale.

The lift (due to pressure differences etc) needs a reaction force (required by Newton). You can't have one without the other. You can calculate the lift of a wing using lift coefficients, air density, velocity etc and lo and behold that force will be balanced by the mass x acceleration of the downward airflow. You can't have one without the other - the lift has to "push" against something.

You can't claim that significant amounts of air aren't directed downwards behind wings. Have you seen wingtip vortices in cloud? Could a helicopter hover without a downdraft? The wingtips vortices have an overall downward movement - planes can't fly without them.

http://www.av8n.com/irro/profilo1_e.html
http://www.av8n.com/how/htm/airfoils.html#sec-circ ulation-vortices
http://amasci.com/wing/whyhard.html
http://www.diam.unige.it/~irro/gallery.html

There is no one single effect that causes a wing to produce lift - it is a combination of interrelated effects:
http://www.av8n.com/how/htm/airfoils.html#sec-cons istent

a similar project is babybot. Short extract: Our scientific goal is that of uncovering the mechanisms of the functioning of the brain by building physical models of the neural control and cognitive structures. In our intendment physical model are embodied artificial systems that freely interact in a not too unconstrained environment. Also, our approach derives from studies of human sensorimotor and cognitive development with the aim of investigating if a developmental approach to building intelligent systems may offer new insight on aspects of human behavior and new tools for the implementation of complex, artificial systems. (BTW: that project has been around since 2000.... )

OK.
This article/summary mess up is really ridiculious.

Just a few quick point ignoring corrections people have made already.
100 us is too much latency for gigE NIC -> NIC uncontested LAN latency.
It was 50 us in 1999 without the amazing other work thats been done since.

Here's a more fair comparision ...

http://www.disi.unige.it/project/GAMMA/

10 us (roundtrip though a switch)
6 us (direct)

Ok, so 200 ns still wins, but at least be honest.
There are a lot of complex tradeoffs here, but the
point is the article is a little slanted, IMO.

So a real group would:
  - price possible interconnects
  - Cpu/Node costs.
  - per-node switching fabric costs.

At maximize performance per $ for the money they have.

So there choice depend on:
1) How much money they have.
2) How what your code looks like.

saying everyone _needs_ these special interconnect for high speed is wrong
in fact I think most groups would be better buying more nodes or NICs.

- Garick

LOL When "babybot" goes to grap the ball watch how fast he gets his hand out of the way!

Obviously babybot doesn't know it's own strength! LOL

Main website

GAMMA is a high-performance, low-latency, non-routable protocol, designed for clusters on a single LAN. The web page isn't that well written, but the code itself seems prety solid and is for the 2.6 kernels.

You obviously have no clue as to what Infiniband is or is capable of. First off 4x Infiniband is 10 times faster then Ethernet at 10 gigabits/sec.

4X infiniband is 10 Gb/s signal rate but actually 8 Gb/s data rate (8b/10b encoding). This is one of many facts that the IB marketing dept. keep forgetting (I keep telling them, but they won't listen for some reasons).

GigE and TCP are quite inefficient when compared to Infiniband

TCP over Infiniband is as inefficient, it has nothing to do with GigE. People use IP over GigE because it's convenient, but you can use GigE without IP if you talk directly to the hardware. Some have tried and are still trying http://www.disi.unige.it/project/gamma/, but the main problem is the lack of hardware documentation from GigE vendors and the short life span of GigE chips.

I even read that a 1024 node cluster using GbE was just as fast as a 256 node cluster using IB

It's interesting to note that there are not many 256 nodes clusters in production with IB at the moment, even less with 1024 nodes. Second, just as fast doing what ? A pointless benchmark specially tuned for Infiniband as the IB supporters are used to publish or real-world applications ? Yes, high speed interconnects make a difference but GigE is just fine for a lot of the HPC applications I have seen so far.

So before you start talking out of your ass do some research like I did.

Don't believe everything you read, and don't drink the cool-aid that fast. Look at the Top500 just to see what machines are out there, not for the ranking (Linkpack is useless). You will see that there are quite a lot of GigE clusters and not that many IB ones. It's a matter of economics: if IB makes sense, people will buy it. These days, they buy much more GigE (or other) than IB.

the University of Genova (Italy) has a project on a baby-robot, that is quite more advanced than that

Your are obviosly wrong if you think that MiniMax alone, even coupled with processing power of DeepBlue could achieve anything. Read here about other factors involved.

Hey, it even says so, in bold: Stating that the fluid flowing above the airfoil is accelerated with respect to the fluid flowing below it ``because it must travel for a longer route in the same time'' is then definitely wrong. Betrayed but your own reference texts, eh?

As harlows_monkey says, in order to understand why the streams do meet if there is a correct angle of attack, you do need deeper insight into aerodynamics than is spelled out in the simple "lay-man's" explanation.

>Anyone see the problem with that? The first
>problem is that no reason is given for the
>airstream over the top to have to meet up
>with the airstreem under the bottom. Why
>can't it just flow straight back?

See here for one of many explanations of the Kutta condition, one of the foundational principles of aerodynamics. This has nothing to do with an explanation for the layman. Basically, it states that the air MUST meet smoothly at the back of the wing.

Logically, if you spend some time thinking about the flow, you cannot possibly construct a situation where the air above the wing somehow slips past the air below. Remember that a jet moves so fast that its wing is only passing through a portion of the air for fractions of a second - it's simply not possible to make the air move fast enough to slip like this.

This principle has been demonstrated NUMEROUS times. You can demonstrate it very easily with a line of smoke through which a wing passes, among a zillion other simple experiments.

Apple's Cocoa or cocoa, the computation in commutative algebra system?

I spent many frustrating days recently downoading, compiling/porting and testing just about every freely-downloadable NN package I could get my hands on, for GNU/Linux, Win32 and Digital Unix -- just about everything that claimed to be a backprop simulator. Many were buggy or difficult to get up and running. Others were difficult to adapt or embed in a real application. Many didn't support a batch mode, or even worse, proper stopping conditions for training. Three clean, stable packages I found were: Aspirine/Migraines, Nevprop, and BPNN, in order of program complexity from most to least complex. However these programs had their limitations too, IMO. I just needed something robust, embeddable, reliable and fast for a University project.

The best software I found, which I cannot recommend highly enough, is MBP (Matrix BackPropagation) by Davide Anguista, available at ftp://ftp.esng.dibe.unige.it/neural/MBP/ . This uses highly-optimized matrix multiplication techniques for both forward and backward propagation, and the accompanying paper is worth a read just to see how the guy got his matrix multiplication so fast. MBP as a backprop simulator is 20-30 times faster than standard backprop NNs. And the code is clean, well-written, highly portable and stable. I take my hat off to Mr. Anguista.