VnutZ Domain
Copyright © 1996 - 2020 [Matthew Vea] - All Rights Reserved

2015-02-01
Featured Article

Genetic Hardware Evolution

[index] [662 page views]
Tagged As: Algorithms, Science, and Technology

Genetic Algorithms, "are a way of solving problems by mimicking the same processes mother nature uses. They use the same combination of selection, recombination and mutation to evolve a solution to a problem." They've often been used in software but have largely not been applied to hardware. Enter Dr. Adrian Thompson and some 100 cell FPGAs. He created an experiment where a computer would analyze the evolution of an FPGA configuration over time until it could reliably differentiate between two tones (and later two voice commands). After nearly 4000 evolutions, Dr. Thompson's creation demonstrated a peculiar quirk of the process - unexplainable hardware nuances themselves became part of the design. For instance, the final candidate included some logical loops that were never actually part of the circuit - but removing them results in the chip's failure. More so, putting the same configuration onto another chip resulted in failure. The algorithm literally adapted to magnetic flux, a "gray area" between transistor ON/OFF states, and other performance quirks of its native FPGA host in order to fulfill its task.



More site content that might interest you:

You just can't make this stuff up ... but supporters don't think any of it is a problem.


Try your hand at fate and use the site's continuously updating statistical analysis of the MegaMillions and PowerBall lotteries to choose "smarter" number. Remember, you don't have to win the jackpot to win money from the lottery!


Tired of social media sites mining all your data? Try a private, auto-deleting message bulletin board.