Cla[mp]ping
March 22nd, 2006 at 05:27 pm
An interesting keynote this morning. Perhaps this is old news to the VI-<3 folks, but see Section IV on the 11th page (numbered 177) of this paper by Thompson et al.
And the problem with energized, engaging speakers is that their presentations seem to the listener to go by so much faster than soporific, unintelligible speakers.
And the problem with the latter is that there are so many of them.
And note that I didn’t say “interesting” or “uninteresting.” The ‘interestingness’ of a talk’s content is orthogonal to how well it’s presented.
maxg
Both comments and trackbacks are closed.
Yup those VI-<3 types hear about all the cool stuff first. Actually the first person to tell me about this was Varady so for a while I thought it was a lie but then I heard it somewhere else. Actually I considered trying to do something similar for the 6.111 project and then realized that I know nothing about genetic algorythms and the half hour it took to load those FPGAs might slow the process down a bit. It’s a very cool idea. And I’d be very interested to see what circuit it came up with other than the lines between functional blocks, but as they touched on, the serious downside is reliability. They did look at temperature variation which was a problem. They did not say anything about putting it on a different FPGA. And the fact that unconnected blocks are important for operation says to me that you better not have anything else running on that FPGA… (kinda defeats the point of using such a small portion of it).
So very cool stuff but I haven’t been eliminated quite yet (or even those silly digital types - well, at least by computers). I understand the idea behind using the actual FPGA but I wonder if it would be better to put it in a simulator where you could run each circuit against a number of conditions (standard analog design practice is run the circuit at fast, normal and slow conditions and then also throw in a little noise here or there).
Sounds like an interesting conference - I look forward to hearing more.
David | March 23rd, 2006 at 09:27 pm
Ok after ripping on genetically designed circuits I do have to admit that I have no idea how to design something that distinguishes between 1 kHz and 5 kHz on an FPGA without using a very large counter or delay.
David | March 23rd, 2006 at 10:35 pm
According to the speaker, the circuit worked only on this 10×10 block of this FPGA. Not in simulation, not on another array. He had a circuit diagram, with the best part being a long loop at the end where the output is fed as feedback through an even number of NOT gates back to the output. Of course building the circuit from standard components gets you nothing.
But the point is: clearly, a circuit capable of exploiting subtle electromagnetic effects to produce useful behavior under limited functional and space resources must have been a product of the Great Circuit Designer in the sky and His infinite wisdom.
Or just mutation and a fitness function.
maxg | March 27th, 2006 at 02:31 pm
What?! - How did you find out about the Great Circuit Designer in the Sky? Next thing you know everyone will know the proper way to sacrifice the great capacitor for its magic smoke and I’ll really be out of a job!
Mutation and a fitness function - ha - there’s no way genetics and evolution could create an electric system able to handle situations that humans have had trouble designing systems for. Nope I sure can’t think of anything like that.
David | March 27th, 2006 at 11:31 pm