Memcomputing NP-complete problems in polynomial time using polynomial resources and collective states

[Ethan Waldo]

Interesting pictures and descriptions of memprocessor network:



http://advances.sciencemag.org/content/advances/suppl/2015/06/30/1.6.e1500031.DC1/1500031_SM.pdf

________________________________
From: Tacos <tacos-bounces+ewaldo=healthetechs.com at amrad.org> on behalf of Ethan Waldo <ewaldo at healthetechs.com>
Sent: Monday, July 06, 2015 12:45 PM
To: AMRAD
Subject: Memcomputing NP-complete problems in polynomial time using polynomial resources and collective states


I found {http://advances.sciencemag.org/content/1/6/e1500031.full} very interesting.  It describes a "memcomputing" system that behaves more like the brain and allows storing and processing of information simultaneously, very useful for machine learning or any other data heavy processing applications.  In big data processing via a "memcomputer", problems could scale linearly with linearly increasing data rather than current computing which often increases exponentially with linearly increasing data.  Such a feat has only been theoretically possible with quantum computing.



I thought the article might be of interest here because the algorithms look suspiciously like an RF application especially since this currently operates in the analog domain (researchers are looking to make this digitial....ADC?).  The article also describes using frequency spectrum, harmonics, and Fourier transforms to associate the collective state.  I am certainly no scientific or RF expert.



Even if they don't figure out an efficient way to make this digital, I could definitely see analog compute engines being developed and bundled in to an add-on card, similar to CUDA processing on GPUs.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.amrad.org/pipermail/tacos/attachments/20150706/020961df/attachment.html>