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

Digital?
By LRonaldHubbs on 7/25/2013 7:17:22 AM , Rating: 2
quote:
In a neuromorphic chip, neurons are modeled as digital circuits, such as the one pictured.

That is not a digital circuit. The ouput may be digital, but the presence of capacitors and current mirrors makes this an analog-digital hybrid circuit, aka mixed-signal.




RE: Digital?
By LRonaldHubbs on 7/25/2013 7:18:38 AM , Rating: 2
Note that this was intended as clarification and not criticism. I wouldn't expect anyone but an EE to have known the difference.


RE: Digital?
By Monkey's Uncle on 7/25/2013 10:27:14 AM , Rating: 2
So you are saying that when I look at the motherboard of my computer and see a couple hundred capacitors on it that it is not a digital circuit?

Sorry, but even a DRAM memory cell requires a capacitor to work and DRAM memory is about as 'digital' as circuits get.


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.


RE: Digital?
By mik123 on 7/25/2013 12:18:34 PM , Rating: 2
By your logic, there's no such thing as "digital circuit".

The difference between "digital" and "analog" circuits is the way we interpret their input/output signals. If we only care about the specific ranges those signals fall into, rather than their precise values, then the signals are digital.

If both input and output signals are digital, then the circuit is digital. If input is digital, and output is analog, or vice versa, then the circuit is mixed signal (for example, ADC or DAC). If both input and output signals are analog, then the circuit is analog.

Presence of capacitors, or any particular physical implementation, has nothing to do with a circuit being digital or analog. You can build a digital circuit with wooden sticks, or an analog circuit with transistors.
It's all about the signal you are interested in.


RE: Digital?
By LRonaldHubbs on 7/25/2013 2:07:15 PM , Rating: 3
quote:
By your logic, there's no such thing as "digital circuit".

My statements imply no such thing. Of course there is such a thing as a digital circuit. It's a matter of scope. I said that all digital circuits are analog if you look at a low enough level, and that is 100% true. I never said that such circuits weren't still digital from a high-level perspective.

If you want to treat the circuit in this article as a black box, and use it as a component to build some other circuit, you certainly could build something which is digital in nature. That is the direction that you and Monkey'sUncle are trying to take this discussion, but that is not what my OP was talking about. All my OP was saying is that the contents of the black box are not digital. That circuit, as drawn in the image above, contains both analog and digital elements, and that means it is mixed-signal.

quote:
The difference between "digital" and "analog" circuits is the way we interpret their input/output signals. If we only care about the specific ranges those signals fall into, rather than their precise values, then the signals are digital. If both input and output signals are digital, then the circuit is digital. If input is digital, and output is analog, or vice versa, then the circuit is mixed signal (for example, ADC or DAC). If both input and output signals are analog, then the circuit is analog.

These distinctions/definitions are your own. This is not how we distinguish between analog and digital circuits in industry.

quote:
Presence of capacitors, or any particular physical implementation, has nothing to do with a circuit being digital or analog.

The presence of capacitors or other analog components does have something to do with it. Here are two key reasons why this is true, and also why the circuit above is not digital.

1) A digital circuit only has 2 internal states: 0 or 1. Yes you can also have tri-state logic where Z is also possible, but if you are talking about Boolean logic then there are only two states. If a circuit contains only pullup and pulldown structures and every node can be described by a Boolean state, then it is digital. If a circuit contains structures which produce variable levels that are not Boolean states, then it is analog. A capacitor is an analog device because it charges or discharges over time and its voltage level is continuously variable across the entire applied operating range (not just discrete Boolean states). Take a look at the waveform in the schematic right above Cmem. That is not a digital waveform.

2) Analog semiconductor circuits are fundamentally different from digital ones in the way that they are fabricated. For any particular manufacturing process/node, the FETs used for analog circuitry are structurally and visually different from logic FETs. The cricuit above contains two current mirrors. These structures rely on the two involved FETs being very well matched and for this reason are constructed from analog FETs, not logic FETs. If you wanted to fabricate this circuit, the mask layout would contain distinct analog and digital device regions. That makes it a mixed-signal circuit.

quote:
You can build a digital circuit with wooden sticks, or an analog circuit with transistors.

I see what you're getting at, but you're stumbling on terminology. You could build a system from sticks whose outputs (whatever they happen to be) are digital in nature, but that system would not be a circuit in any traditional sense, nor would its internal function be digital.


RE: Digital?
By mik123 on 7/25/2013 3:15:02 PM , Rating: 2
Let's look at the waveform in the schematics above Cmem node. You call it "not a digital waveform". Perhaps this is the source of confusion.

That waveform, as shown, can be interpreted as "digital", or "analog", depending on what that signal means to us. Is this signal intended to trigger something? Or are we interested in all of its possible values in a range? If your job is to design a capacitor for that circuit, you might well be interested in the continuous values of that signal. For you, it's analog. However, for someone who is interested in input/output of that circuit, that waveform is digital - it's either triggers the FETs down the wire, or it does not, which is all that matters. Same with DRAM cells - if you design them, they are analog circuits to you, however, if you just use them, they are digital circuits: for a user, a signal that passes through a DRAM cell is 100% digital.

You mention analog and digital FETs, yes, sure they are fabricated differently. But that is not their fundamental difference. The fundamental difference is in how we interpret the signals that flow through them. In one case, the signal is analog, in other it's digital.
My point is, that any element can be used as a digital circuit, or as an analog circuit. If you take a FET which was designed for analog operation, and use it as on on/off switch, then this FET is a digital device (or a simple digital circuit, if you will. It's not how it's built, it's what it does is important! The way we interpret the signal is what makes something "analog" or "digital", be it an element, a circuit, or the entire system.

In my opinion, the "signal perspective" is a more clear way to explain "digital vs analog" concept to a novice. By the way, this is how it's explained in most textbooks. See, for example, the explanation on Wikipedia:
http://en.wikipedia.org/wiki/Digital_circuits

It's all about signals, and the information they represent.


RE: Digital?
By Sertis on 7/25/2013 6:42:18 PM , Rating: 2
It could be that it was taken out of context, the image is an example of a circuit that appears to be used for continuous analog signal processing, while the paper referenced discusses using purely digital logic to achieve better density and lower power consumption. The image is more of an example of "how neurological circuits used to be".


RE: Digital?
By LRonaldHubbs on 7/26/2013 7:09:54 AM , Rating: 2
Good point, it may well be taken out of context. I haven't had time to go read the source publication yet.


RE: Digital?
By LRonaldHubbs on 7/26/2013 7:02:11 AM , Rating: 2
This entire post exactly demonstrates the disconnect that I explained in my second paragraph above. You want to treat this circuit as a black box and say that because the outputs are digital, and downstream circuits see it as digital, the box must also be digital. By contrast, I am saying that because the picture above is NOT a black box, and we have been presented the exact contents of the box, the caption for that picture should reflect the contents; those contents are mixed-signal. Also, the inputs to this circuit are analog (current and voltage sources, not digital signals), so by your own definition this must be a mixed-signal circuit.

You do seem to acknowledge that from a circuit designer's perspective, which is indeed my perspective, this circuit is not digital. Because we are being shown the schematic for this circuit by itself, not in some greater context, that is exactly the perspective which should be used to describe it. If, on the other hand, we were shown a picture of a higher-level system view in which this circuit interfaces to other circuits, then by all means take the user's perspective and call it a digital system.


RE: Digital?
By mik123 on 7/26/2013 3:13:30 PM , Rating: 2
I've never called this circuit "digital". My point is that it does not make sense to call any circuit "digital", "analog", or "mixed signal" until you know what kind of signal flows through the circuit.
That's why I disagreed with your original statement that this circuit can't be digital just because it has a capacitor.

Just to be clear, what is your definition of a digital or analog circuit?

Your argument about a black box is not relevant, because I am looking inside the circuit. However, what I'm looking for is the signal type, not the type of elements, or circuit design. If I am looking inside a DRAM cell, and the signal there appears digital, from my perspective, then the DRAM cell is a digital circuit to me. Because, from my perspective, DRAM cell takes a digital signal as input, stores a digital signal, and outputs a digital signal. If you can show to me that a signal inside a DRAM becomes analog, then I will immediately agree that a DRAM cell is not a pure digital circuit. What's relevant here is, again, the nature of the signal, not the design of a circuit. Do you see my point?


RE: Digital?
By LRonaldHubbs on 8/8/2013 11:26:14 AM , Rating: 2
quote:
I've never called this circuit "digital".

You did implicity by disgreeing with the assertion that it is not digital.

quote:
My point is that it does not make sense to call any circuit "digital", "analog", or "mixed signal" until you know what kind of signal flows through the circuit.

But we do know this information because the schematic was given to us. The inputs are voltage and current sources, not digital signals.

quote:
That's why I disagreed with your original statement that this circuit can't be digital just because it has a capacitor.

That wasn't the only reason that I gave in my original statement. The circuit also contains current mirrors, which are classic analog structures.

quote:
Your argument about a black box is not relevant, because I am looking inside the circuit.

It's not irrelevant. You said that to determine if this circuit is digital we should consider the input and output signal types and look at how upstream and downstream circuits see this circuit. That by definition is the black-box approach.

quote:
"If I am looking inside a DRAM cell, and the signal there appears digital, from my perspective, then the DRAM cell is a digital circuit to me."

The DRAM cell node leaks over time, and if it is not refreshed it will become indeterminate. That makes it analog, not digital. The DRAM cell's state is "read" by another analog circuit called a sense amplifier. The read data output from the sense amp the absolute lowest level of a DRAM which can be considered digital from an system I/O perspective.

quote:
What's relevant here is, again, the nature of the signal, not the design of a circuit.

Like I've said repeatedly, if you want to take the system POV and look only at I/O signals then this is true. If, on the other hand, you're looking inside the black box and taking the circuit designer POV, then the internal nodes of the circuit are relevant.

quote:
Do you see my point?

I see your point when taking the system POV, but I clearly communicated that because a schematic is given, the caption for that schematic should take the circuit designer POV. I don't understand why you're continuing to push the system POV while ignoring the context of both my posts and the image in question.

quote:
Just to be clear, what is your definition of a digital or analog circuit?

From a system POV, treating all circuits as black boxes:
digital circuit -- digital inputs and ouputs
analog circuit -- analog inputs and outputs
mixed signal -- some combination of the two

From what I can tell, ^that^ is exactly your defintion.

From a circuit designer perspective, looking at the contents of the box:
digital circuit -- all internal nodes can be represented by Boolean expressions
analog circuit -- no nodes have Boolean expressions
mixed-signal -- anything in between the two

My point from the very beginning has been that because we were presented the circuit a standalone unit and given an image of its exact contents, the caption for that image must assume the circuit designer's perspective.


RE: Digital?
By LRonaldHubbs on 7/26/2013 7:09:08 AM , Rating: 2
I meant to have a third bullet point in there but couldn't remember it by the time I got that far along.

3) If you wanted to simulate this circuit, you could not [easily] do so with a digital simulator like Verilog. Maybe if you already understood the circuit very well you could adequately fake it out with some clever use of tristate logic and drive strengths, but internally your model would not be true to the actual circuit. There is no way any CAD tool is going to produce a usable Verilog netlist from this schematic. In order to truly characterize this circuit you would instead have to use transient simulation, and that means that it is not digital.


RE: Digital?
By mik123 on 7/26/2013 3:49:50 PM , Rating: 2
This is not a relevant argument. If the input and output signals to this circuit were digital, then you would be able to describe the circuit behavior with Verilog. That's all that matters to Verilog - the signal have to be digital, it does not care about the circuit (it's not a circuit simulator anyway).


RE: Digital?
By ranran on 7/31/2013 1:59:28 PM , Rating: 2
To [LRonaldHubbs] & [mik123]:

Gentlemen, While this discussion is certainly out of my area of expertise, I must congratulate you two for having a serious, respectful, and mature debate on this subject without resorting to infantile name-calling or other ridiculous behaviour.

It is refreshing to be able to read comments and not have to skim through "garbage-talk".


RE: Digital?
By LRonaldHubbs on 8/8/2013 11:32:00 AM , Rating: 2
Right, but the inputs to this circuit are not digital, they are voltage and current sources. So again, this thing can't [easily] be modelled with Verilog.

quote:
it does not care about the circuit (it's not a circuit simulator anyway).

It certainly is a circuit simulator if you feed it a transistor-level structural netlist.


RE: Digital?
By mik123 on 7/25/2013 12:20:27 PM , Rating: 2
Sorry, I was replying to LRonaldHubbs


RE: Digital?
By DennisB on 7/25/2013 7:46:09 AM , Rating: 2
I'm not sure about the "new" thing here or the difference to the efforts in neuro processors for the past 3+ decades. Any idea?
Some papers from the past:
http://www.ieeeexplore.net/xpl/articleDetails.jsp?...
http://www.ieeeexplore.net/xpl/articleDetails.jsp?...


RE: Digital?
By LRonaldHubbs on 7/25/2013 8:22:59 AM , Rating: 2
This snippet from the article summarizes the novelty:
quote:
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.


Basically the designers of this circuit traded off functionality for density, enabling them to put many more 'neurons' onto a chip.

For reference, here is what's IBM has done:
http://www.modha.org/papers/013.CICC2.pdf (see Figure 3)

and University of Manchester:
http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumb... (see Figure 1)


"My sex life is pretty good" -- Steve Jobs' random musings during the 2010 D8 conference














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