3 ms·

Researchers make a “diffusive” memristor that emulates how a real synapse work

72 points by yexponential 10y ago

ChuckMcM 10y ago
It is an interesting result but until it is connected with a system that can implement the other parts of learning we're left with a model of a neuron. Back in the 90's when neural networks were the big thing the first time people built what they considered to be very accurate neuron models connected together into networks as a way of building a system. While this gives you a way to do that in hardware, and so potentially much faster and denser than the software systems, the missing bit is the system when connects these things together and feeds them inputs and pulls off outputs such that the system can be trained. Still looking for that paper.
- adrusi 10y ago
  Could a network be trained in software on powerful expensive hardware and then programmed onto some kind of neural FPGA that uses these memresistors to be used in power/space constrained systems?
  - cbennett 10y ago
    Short answer: yes. Basically there are two options for future ReRAM (memristive) learning system: in-situ/on-chip learning , in which all learning rules are locally derived and enforced, and ex-situ learning systems in which we do what you suggested- import weights from more computationally/power expensive substrates. there is probably abundant promise in both approaches moving forward. I recommend looking at some recent papers by the Strukov group [1] as well as my own [2] to see the limitations of these approaches. Strukov paper skirts around the issue to a certain degree but they admit in supplementary material scaling issues are not favorable with their approach. our work takes the 'neural FPGA' approach quite literally. But, their approach may , with some improvements , do rather well for an on-chip backprop implementation. Let's see what they do next. Lastly, as far as hybrid approaches, there is a recent IBM paper which is really nice which talks about deep neural net acceleration with ReRAM. If you're really curious let me know and I"ll try to dig it up. [1] http://www.nature.com/nature/journal/v521/n7550/abs/nature14441.html http://www.nature.com/nature/journal/v521/n7550/abs/nature14... [2]http://www.nature.com/articles/srep31932 http://www.nature.com/articles/srep31932
    - maxjohansen 10y ago
      I would love to hear more about your research, please try to dig up the links.
      - cbennett 10y ago
        Glad to hear you find it exciting. I do too.. its a really hot field at the moment and I mean that in the good, not bad way ;) Lots of groups working in parallel on somewhat orthogonal design and architecture issues, with a variety of different considered devices, but a common basis set is emerging ;) So, here's the paper I mentioned above. I think this is very methodical and inventive and definitely one of the best yet at considering confluence of DNNs and memristive (ReRAM) devices. A quick search revealed this was already on HN. https://arxiv.org/abs/1603.07341 https://arxiv.org/abs/1603.07341 So, I already mentioned the iconic Strukov paper above and my own which is really quite similar to Strukov in learning strategy/philosophy, except for we used entirely chemical and 'slow' devices , which may be quite interesting for brain emulation. (remember the brain operates in the mS , or microsecond regime and not nanosecond). Here's another article I just stumbled upon a few days ago but which looks quite promising and brings us into the territory of a more un-supervised /probabilistic algorithm for learning. http://www.nature.com/articles/ncomms12611 http://www.nature.com/articles/ncomms12611
        p1esk 10y ago
        What do you think about "chip in the loop" approach for training, which was popular in the 90s for hardware NN implementations?
    - zamalek 10y ago
      For a simple multi-layer feed-forward network, couldn't you just just use "classical" components - seeing as the parameters of the network would never change (having been trained beforehand)? I.e. Would you actually need memristors?
sverige 10y ago
Memristors are fascinating. Here's my question: How much do biologists understand about how synaptic systems work in living organisms? In other words, is this more likely to be helpful to people studying how memories and thoughts are created and retained in living organisms, or is this more likely to be helpful to people studying how to make artificial systems behave more like organic systems?
- Karlozkiller 10y ago
  Things like memories are not well understood from what I've gathered as a Master's Student in cognitive science. I mean, it is thought to be stored in the form of persistent patterns in neurons and we know certain areas of the brain that are vital for the formation of memories. But there is no clear model of how memories are formed. There has been rapid increase in understanding of such things lately and I would say neural networks give some insight and ways to explore and experiment with different configurations further. A physical model like this could probably serve as a proof of concept and maybe help further knowledge in the are, but I suspect computerized simulations will serve this purpose better. tldr; as much of a layman I believe we know too little about the brain to make a proper physical brain.
  - posterboy 10y ago
    I saw pictures labeled extracted from live brain scans. How that was done I don't know, maybe alpha or beta waves. At least that's short term memory.
    - Karlozkiller 10y ago
      Brain scans are not as accurate as one might think. Most or all methods merely show activation of neurons or groups of neurons over a timescale. Typically you give a person, say a memory task, you look at which areas fire up as he tries to memorise a word list and from that you extrapolate whatever you can from it. I also believe working memory and short term memory are to some extent better understood than long term memory. actually the least known process regarding memory I think is how memories pass from working memory or short term memory into long term memory.
pohl 10y ago
I was surprised today that the entry on Wikipedia still qualifies the memristor as "hypothetical", and says that "there are...some serious doubts as to whether the memristor can actually exist in physical reality". I thought maybe research was farther along than that. https://en.wikipedia.org/wiki/Memristor https://en.wikipedia.org/wiki/Memristor
- cbennett 10y ago
  In truth it is. To practitioners in the field building circuits or doing simulations with nanodevices, this is somewhat of a tiresome debate. To give some background, Leon Chua made certain claims about a hypothetical fourth circuit element and these debates largely stem back to claims about circuit analysis and mathematics. Basically his models predict a perfect device which, to my knowledge , has not been experimentally realized (to the contrary of HP's claims). However , the funny thing is it doesn't really matter. We don't need a perfect memristor to build interesting and useful nanoionic and nano-redox circuits performing non-linear computational tasks. As modelers though, we do need to be careful making ideal claims about eternal non-volatility and device life (of course). Many point out that to the contrary of being ideal, these devices are extremely variable and imperfect- which is true. Anything built using nanofab techniques at the academic level (excluding semi-con industrial processes) will be.. Btw, If you want more physics depth on this , I can recommend any paper or book by Waser. They are all good. http://eu.wiley.com/WileyCDA/WileyTitle/productCd-3527334173.html http://eu.wiley.com/WileyCDA/WileyTitle/productCd-3527334173... Edit: adding a link to the first book chapter of aforementioned book I found which is already rather good. https://application.wiley-vch.de/books/sample/3527334173_c01.pdf https://application.wiley-vch.de/books/sample/3527334173_c01...
- hodwik 10y ago
  I was confused by that as well. I was under the impression HP made working memristors years ago.

72 points by yexponential 10y ago

It is an interesting result but until it is connected with a system that can implement the other parts of learning we're left with a model of a neuron. Back in the 90's when neural networks were the big thing the first time people built what they considered to be very accurate neuron models connected together into networks as a way of building a system. While this gives you a way to do that in hardware, and so potentially much faster and denser than the software systems, the missing bit is the system when connects these things together and feeds them inputs and pulls off outputs such that the system can be trained. Still looking for that paper.

adrusi 10y ago
Could a network be trained in software on powerful expensive hardware and then programmed onto some kind of neural FPGA that uses these memresistors to be used in power/space constrained systems?
- cbennett 10y ago
  Short answer: yes. Basically there are two options for future ReRAM (memristive) learning system: in-situ/on-chip learning , in which all learning rules are locally derived and enforced, and ex-situ learning systems in which we do what you suggested- import weights from more computationally/power expensive substrates. there is probably abundant promise in both approaches moving forward. I recommend looking at some recent papers by the Strukov group [1] as well as my own [2] to see the limitations of these approaches. Strukov paper skirts around the issue to a certain degree but they admit in supplementary material scaling issues are not favorable with their approach. our work takes the 'neural FPGA' approach quite literally. But, their approach may , with some improvements , do rather well for an on-chip backprop implementation. Let's see what they do next. Lastly, as far as hybrid approaches, there is a recent IBM paper which is really nice which talks about deep neural net acceleration with ReRAM. If you're really curious let me know and I"ll try to dig it up. [1] http://www.nature.com/nature/journal/v521/n7550/abs/nature14441.html http://www.nature.com/nature/journal/v521/n7550/abs/nature14... [2]http://www.nature.com/articles/srep31932 http://www.nature.com/articles/srep31932
  - maxjohansen 10y ago
    I would love to hear more about your research, please try to dig up the links.
    - cbennett 10y ago
      Glad to hear you find it exciting. I do too.. its a really hot field at the moment and I mean that in the good, not bad way ;) Lots of groups working in parallel on somewhat orthogonal design and architecture issues, with a variety of different considered devices, but a common basis set is emerging ;) So, here's the paper I mentioned above. I think this is very methodical and inventive and definitely one of the best yet at considering confluence of DNNs and memristive (ReRAM) devices. A quick search revealed this was already on HN. https://arxiv.org/abs/1603.07341 https://arxiv.org/abs/1603.07341 So, I already mentioned the iconic Strukov paper above and my own which is really quite similar to Strukov in learning strategy/philosophy, except for we used entirely chemical and 'slow' devices , which may be quite interesting for brain emulation. (remember the brain operates in the mS , or microsecond regime and not nanosecond). Here's another article I just stumbled upon a few days ago but which looks quite promising and brings us into the territory of a more un-supervised /probabilistic algorithm for learning. http://www.nature.com/articles/ncomms12611 http://www.nature.com/articles/ncomms12611
      - p1esk 10y ago
        What do you think about "chip in the loop" approach for training, which was popular in the 90s for hardware NN implementations?
  - zamalek 10y ago
    For a simple multi-layer feed-forward network, couldn't you just just use "classical" components - seeing as the parameters of the network would never change (having been trained beforehand)? I.e. Would you actually need memristors?

sverige 10y ago

Memristors are fascinating. Here's my question: How much do biologists understand about how synaptic systems work in living organisms? In other words, is this more likely to be helpful to people studying how memories and thoughts are created and retained in living organisms, or is this more likely to be helpful to people studying how to make artificial systems behave more like organic systems?

Karlozkiller 10y ago
Things like memories are not well understood from what I've gathered as a Master's Student in cognitive science. I mean, it is thought to be stored in the form of persistent patterns in neurons and we know certain areas of the brain that are vital for the formation of memories. But there is no clear model of how memories are formed. There has been rapid increase in understanding of such things lately and I would say neural networks give some insight and ways to explore and experiment with different configurations further. A physical model like this could probably serve as a proof of concept and maybe help further knowledge in the are, but I suspect computerized simulations will serve this purpose better. tldr; as much of a layman I believe we know too little about the brain to make a proper physical brain.
- posterboy 10y ago
  I saw pictures labeled extracted from live brain scans. How that was done I don't know, maybe alpha or beta waves. At least that's short term memory.
  - Karlozkiller 10y ago
    Brain scans are not as accurate as one might think. Most or all methods merely show activation of neurons or groups of neurons over a timescale. Typically you give a person, say a memory task, you look at which areas fire up as he tries to memorise a word list and from that you extrapolate whatever you can from it. I also believe working memory and short term memory are to some extent better understood than long term memory. actually the least known process regarding memory I think is how memories pass from working memory or short term memory into long term memory.

pohl 10y ago

I was surprised today that the entry on Wikipedia still qualifies the memristor as "hypothetical", and says that "there are...some serious doubts as to whether the memristor can actually exist in physical reality". I thought maybe research was farther along than that. https://en.wikipedia.org/wiki/Memristor https://en.wikipedia.org/wiki/Memristor

cbennett 10y ago
In truth it is. To practitioners in the field building circuits or doing simulations with nanodevices, this is somewhat of a tiresome debate. To give some background, Leon Chua made certain claims about a hypothetical fourth circuit element and these debates largely stem back to claims about circuit analysis and mathematics. Basically his models predict a perfect device which, to my knowledge , has not been experimentally realized (to the contrary of HP's claims). However , the funny thing is it doesn't really matter. We don't need a perfect memristor to build interesting and useful nanoionic and nano-redox circuits performing non-linear computational tasks. As modelers though, we do need to be careful making ideal claims about eternal non-volatility and device life (of course). Many point out that to the contrary of being ideal, these devices are extremely variable and imperfect- which is true. Anything built using nanofab techniques at the academic level (excluding semi-con industrial processes) will be.. Btw, If you want more physics depth on this , I can recommend any paper or book by Waser. They are all good. http://eu.wiley.com/WileyCDA/WileyTitle/productCd-3527334173.html http://eu.wiley.com/WileyCDA/WileyTitle/productCd-3527334173... Edit: adding a link to the first book chapter of aforementioned book I found which is already rather good. https://application.wiley-vch.de/books/sample/3527334173_c01.pdf https://application.wiley-vch.de/books/sample/3527334173_c01...
hodwik 10y ago
I was confused by that as well. I was under the impression HP made working memristors years ago.