backtop


Print 26 comment(s) - last by LRonaldHubbs.. on Aug 8 at 11:32 AM


  (Source: TriStar Pictures)
Neuromorphic community continues to advance towards offering analogous behavior to mammal brains

Researchers have long hoped to emulate the way living organisms process and store information -- neural networks -- either in code or via simulations of true neurons on supercomputers.  But the Universität Zürich (Univ. of Zürich) and Eidgenössische Technische Hochschule Zürich (ETH Zürich) are dreaming much bigger imagining creating special circuits that mimic neurons in hardware, not software, allowing the speed necessary to perform incredibly compex tasks such as allowing an intelligent robot to recognize the objects it "sees" via retinal sensors or "hears" via cochlea-mimicking devices.

The researchers begin by using a "neuron" circuit to create masses of untrained neurons using standard very-large-scale integration (VLSI) design techniques.  They then map sensor inputs to neurons' bias voltages creating soft state machine style neural networks.

The resulting networks that can best recall the inputs -- in terms of gain, signal restoration, and multistability -- are then preserved.

The brain-like circuit, which researchers dubbed a "neuromorphic" chip, is used in a demo to perform "real-time context-dependent classification of motion patterns observed by a silicon retina."

brain chipNeuromorphic chips learn and process information faster than software models run on traditional hardware. [Image Source: INI]

Giacomo Indiveri, a professor at the Swiss universities' Institute of Neuroinformatics (INI) comments [press release], "Our goal is to emulate the properties of biological neurons and synapses directly on microchips.  The network connectivity patterns [in our latest work] closely resemble structures that are also found in mammalian brains.  Thanks to our method, neuromorphic chips can be configured for a large class of behavior modes. Our results are pivotal for the development of new brain-inspired technologies."

Retinal implant
The INI's new neuromorphic chip uses a retina-like sensor as a visual input for learning.
[Image Source: GeekInfo]

The INI's work builds on University of Sydney Electrical Engineering Professor Andre van Shaik's 1996 digital neuron model [abstract], which consists of transistors and capacitors attached to various voltage and current sources.  

Neuron circuit
In a neuromorphic chip, neurons are modeled as digital circuits, such as the one pictured.
[Image Source: Neural Networks/Elsevier]

This approach (also known as "spiking neural network" hardware) is different from the analog circuit model first demonstrated by the aforementioned Prof. Rodney Douglas (who at the time was a professor at the University of Oxford, UK) and Misha Mahowald, a California Institute of Technology (CalTech) PhD student, back in 1991.

The advantage of the digital approach is that while it lacks in the detailed reproduction of every facet the neuron's electrical behavior it "coarse grains" its basic operation down to a much smaller circuit, allowing large networks of neurons to be built.

International Business Machines Corp. (IBM) is but one of the large companies looking to productize neuromorphic chips.

A study on the work was published [abstract] in July's early edition of the Proceedings of the National Academy of Sciences (PNAS).  Co-authors of the work include Elisabetta Chicca, a postdoctoral research at the INI who since has moved to the Universität Bielefeld in Germany; INI director Prof. Rodney DouglasUeli Rutishauser, a postodoctoral researcher at Frankfurt, Germany's Max Planck Institute for Brain ResearchEmre Neftci, another postdoc at the INI; and Jonathan Binas, a PhD student at the INI.

Sources: ETH Zürich [press release], PNAS [abstract]



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

RE: Digital?
By LRonaldHubbs on 7/25/2013 11:58:38 AM , Rating: 3
You're talking to someone who designs DRAM for a living. Capacitors are analog devices. The DRAM cell is a textbook example of an analog circuit. The DRAM cell leaks charge over time and requires a mechanism for periodically refreshing its state. If you think that is a characteristic of a digital circuit, then you don't even know what the word means. A DRAM memory chip is only digital because of all the peripheral logic that is wrapped around the array of DRAM cells, logic which was designed for the exact purpose of making it look digital.

Any digital product, including a motherboard, is analog if you look at a low enough level. But we are not talking about motherboards, or any other high-level assembly of smaller circuits. We are talking about the transistor-level circuit which is pictured in this article, and that circuit contains elements which any EE can tell you are analog. That means this is a mixed-signal circuit, not a digital one. Granted, an array of these neuron circuits would indeed behave in a digital manner, but that doesn't mean this circuit is digital.


RE: Digital?
By mik123 on 7/25/2013 12:47:27 PM , Rating: 2
You are confusing people. See my response below.


"A politician stumbles over himself... Then they pick it out. They edit it. He runs the clip, and then he makes a funny face, and the whole audience has a Pavlovian response." -- Joe Scarborough on John Stewart over Jim Cramer














botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki