Memcomputing: a brain-inspired computing paradigm to store and process information on the same physical platform

Date

Oct 2, 2014

Time

1:00 PM - 2:00 PM

Speaker

Massimiliano Di Ventra

Affiliation

Department of Physics, University of California, San Diego, La Jolla, USA

Series

TUD nanoSeminar

Language

en

Main Topic

Materialien

Other Topics

Materialien

Host

G. Cuniberti

Description

I will discuss a novel computing paradigm we named memcomputing inspired by the operation of our own brain. Memcomputing - computing using memory circuit elements or memelements - satisfies important physical requirements: (i) it is intrinsically massively parallel, (ii) its information storing and computing units are physically the same, and (iii) it does not rely on active elements as the main tools of operation. I will then introduce the notion of universal memcomputing machines (UMMs) as a class of general-purpose computing machines based on systems with memory. We have shown that the memory properties of UMMs endow them with universal computing power- they are Turing-complete, intrinsic parallelism, functional polymorphism, and information overhead, namely their collective states can support exponential data compression directly in memory. We have proved that UMMs can solve NP-complete problems in polynomial time, and as an example I will provide the polynomial-time solution of the subset-sum problem when implemented in hardware. Even though we have not proved NP=P, the practical implementation of these UMMs would represent a paradigm shift from present von Neumann architectures bringing us closer to brain-like neural computation. In fact, I will discuss a practical CMOS-compatible realization of this computing paradigm that uses memcapacitors and we have named Dynamic Computing Random Access Memory (DCRAM).

Last modified: Oct 2, 2014, 9:56:37 AM