Tag Archives: neuromorphic computing

China’s neuromorphic chips: Darwin and Tianjic

I believe that China has more than two neuromorphic chips. The two being featured here are the ones for which I was easily able to find information.

The Darwin chip

The first information (that I stumbled across) about China and a neuromorphic chip (Darwin) was in a December 22, 2015 Science China Press news release on EurekAlert,

Artificial Neural Network (ANN) is a type of information processing system based on mimicking the principles of biological brains, and has been broadly applied in application domains such as pattern recognition, automatic control, signal processing, decision support system and artificial intelligence. Spiking Neural Network (SNN) is a type of biologically-inspired ANN that perform information processing based on discrete-time spikes. It is more biologically realistic than classic ANNs, and can potentially achieve much better performance-power ratio. Recently, researchers from Zhejiang University and Hangzhou Dianzi University in Hangzhou, China successfully developed the Darwin Neural Processing Unit (NPU), a neuromorphic hardware co-processor based on Spiking Neural Networks, fabricated by standard CMOS technology.

With the rapid development of the Internet-of-Things and intelligent hardware systems, a variety of intelligent devices are pervasive in today’s society, providing many services and convenience to people’s lives, but they also raise challenges of running complex intelligent algorithms on small devices. Sponsored by the college of Computer science of Zhejiang University, the research group led by Dr. De Ma from Hangzhou Dianzi university and Dr. Xiaolei Zhu from Zhejiang university has developed a co-processor named as Darwin.The Darwin NPU aims to provide hardware acceleration of intelligent algorithms, with target application domain of resource-constrained, low-power small embeddeddevices. It has been fabricated by 180nm standard CMOS process, supporting a maximum of 2048 neurons, more than 4 million synapses and 15 different possible synaptic delays. It is highly configurable, supporting reconfiguration of SNN topology and many parameters of neurons and synapses.Figure 1 shows photos of the die and the prototype development board, which supports input/output in the form of neural spike trains via USB port.

The successful development ofDarwin demonstrates the feasibility of real-time execution of Spiking Neural Networks in resource-constrained embedded systems. It supports flexible configuration of a multitude of parameters of the neural network, hence it can be used to implement different functionalities as configured by the user. Its potential applications include intelligent hardware systems, robotics, brain-computer interfaces, and others.Since it uses spikes for information processing and transmission,similar to biological neural networks, it may be suitable for analysis and processing of biological spiking neural signals, and building brain-computer interface systems by interfacing with animal or human brains. As a prototype application in Brain-Computer Interfaces, Figure 2 [not included here] describes an application example ofrecognizingthe user’s motor imagery intention via real-time decoding of EEG signals, i.e., whether he is thinking of left or right, and using it to control the movement direction of a basketball in the virtual environment. Different from conventional EEG signal analysis algorithms, the input and output to Darwin are both neural spikes: the input is spike trains that encode EEG signals; after processing by the neural network, the output neuron with the highest firing rate is chosen as the classification result.

The most recent development for this chip was announced in a September 2, 2019 Zhejiang University press release (Note: Links have been removed),

The second generation of the Darwin Neural Processing Unit (Darwin NPU 2) as well as its corresponding toolchain and micro-operating system was released in Hangzhou recently. This research was led by Zhejiang University, with Hangzhou Dianzi University and Huawei Central Research Institute participating in the development and algorisms of the chip. The Darwin NPU 2 can be primarily applied to smart Internet of Things (IoT). It can support up to 150,000 neurons and has achieved the largest-scale neurons on a nationwide basis.

The Darwin NPU 2 is fabricated by standard 55nm CMOS technology. Every “neuromorphic” chip is made up of 576 kernels, each of which can support 256 neurons. It contains over 10 million synapses which can construct a powerful brain-inspired computing system.

“A brain-inspired chip can work like the neurons inside a human brain and it is remarkably unique in image recognition, visual and audio comprehension and naturalistic language processing,” said MA De, an associate professor at the College of Computer Science and Technology on the research team.

“In comparison with traditional chips, brain-inspired chips are more adept at processing ambiguous data, say, perception tasks. Another prominent advantage is their low energy consumption. In the process of information transmission, only those neurons that receive and process spikes will be activated while other neurons will stay dormant. In this case, energy consumption can be extremely low,” said Dr. ZHU Xiaolei at the School of Microelectronics.

To cater to the demands for voice business, Huawei Central Research Institute designed an efficient spiking neural network algorithm in accordance with the defining feature of the Darwin NPU 2 architecture, thereby increasing computing speeds and improving recognition accuracy tremendously.

Scientists have developed a host of applications, including gesture recognition, image recognition, voice recognition and decoding of electroencephalogram (EEG) signals, on the Darwin NPU 2 and reduced energy consumption by at least two orders of magnitude.

In comparison with the first generation of the Darwin NPU which was developed in 2015, the Darwin NPU 2 has escalated the number of neurons by two orders of magnitude from 2048 neurons and augmented the flexibility and plasticity of the chip configuration, thus expanding the potential for applications appreciably. The improvement in the brain-inspired chip will bring in its wake the revolution of computer technology and artificial intelligence. At present, the brain-inspired chip adopts a relatively simplified neuron model, but neurons in a real brain are far more sophisticated and many biological mechanisms have yet to be explored by neuroscientists and biologists. It is expected that in the not-too-distant future, a fascinating improvement on the Darwin NPU 2 will come over the horizon.

I haven’t been able to find a recent (i.e., post 2017) research paper featuring Darwin but there is another chip and research on that one was published in July 2019. First, the news.

The Tianjic chip

A July 31, 2019 article in the New York Times by Cade Metz describes the research and offers what seems to be a jaundiced perspective about the field of neuromorphic computing (Note: A link has been removed),

As corporate giants like Ford, G.M. and Waymo struggle to get their self-driving cars on the road, a team of researchers in China is rethinking autonomous transportation using a souped-up bicycle.

This bike can roll over a bump on its own, staying perfectly upright. When the man walking just behind it says “left,” it turns left, angling back in the direction it came.

It also has eyes: It can follow someone jogging several yards ahead, turning each time the person turns. And if it encounters an obstacle, it can swerve to the side, keeping its balance and continuing its pursuit.

… Chinese researchers who built the bike believe it demonstrates the future of computer hardware. It navigates the world with help from what is called a neuromorphic chip, modeled after the human brain.

Here’s a video, released by the researchers, demonstrating the chip’s abilities,

Now back to back to Metz’s July 31, 2019 article (Note: A link has been removed),

The short video did not show the limitations of the bicycle (which presumably tips over occasionally), and even the researchers who built the bike admitted in an email to The Times that the skills on display could be duplicated with existing computer hardware. But in handling all these skills with a neuromorphic processor, the project highlighted the wider effort to achieve new levels of artificial intelligence with novel kinds of chips.

This effort spans myriad start-up companies and academic labs, as well as big-name tech companies like Google, Intel and IBM. And as the Nature paper demonstrates, the movement is gaining significant momentum in China, a country with little experience designing its own computer processors, but which has invested heavily in the idea of an “A.I. chip.”

If you can get past what seems to be a patronizing attitude, there are some good explanations and cogent criticisms in the piece (Metz’s July 31, 2019 article, Note: Links have been removed),

… it faces significant limitations.

A neural network doesn’t really learn on the fly. Engineers train a neural network for a particular task before sending it out into the real world, and it can’t learn without enormous numbers of examples. OpenAI, a San Francisco artificial intelligence lab, recently built a system that could beat the world’s best players at a complex video game called Dota 2. But the system first spent months playing the game against itself, burning through millions of dollars in computing power.

Researchers aim to build systems that can learn skills in a manner similar to the way people do. And that could require new kinds of computer hardware. Dozens of companies and academic labs are now developing chips specifically for training and operating A.I. systems. The most ambitious projects are the neuromorphic processors, including the Tianjic chip under development at Tsinghua University in China.

Such chips are designed to imitate the network of neurons in the brain, not unlike a neural network but with even greater fidelity, at least in theory.

Neuromorphic chips typically include hundreds of thousands of faux neurons, and rather than just processing 1s and 0s, these neurons operate by trading tiny bursts of electrical signals, “firing” or “spiking” only when input signals reach critical thresholds, as biological neurons do.

Tiernan Ray’s August 3, 2019 article about the chip for ZDNet.com offers some thoughtful criticism with a side dish of snark (Note: Links have been removed),

Nature magazine’s cover story [July 31, 2019] is about a Chinese chip [Tianjic chip]that can run traditional deep learning code and also perform “neuromorophic” operations in the same circuitry. The work’s value seems obscured by a lot of hype about “artificial general intelligence” that has no real justification.

The term “artificial general intelligence,” or AGI, doesn’t actually refer to anything, at this point, it is merely a placeholder, a kind of Rorschach Test for people to fill the void with whatever notions they have of what it would mean for a machine to “think” like a person.

Despite that fact, or perhaps because of it, AGI is an ideal marketing term to attach to a lot of efforts in machine learning. Case in point, a research paper featured on the cover of this week’s Nature magazine about a new kind of computer chip developed by researchers at China’s Tsinghua University that could “accelerate the development of AGI,” they claim.

The chip is a strange hybrid of approaches, and is intriguing, but the work leaves unanswered many questions about how it’s made, and how it achieves what researchers claim of it. And some longtime chip observers doubt the impact will be as great as suggested.

“This paper is an example of the good work that China is doing in AI,” says Linley Gwennap, longtime chip-industry observer and principal analyst with chip analysis firm The Linley Group. “But this particular idea isn’t going to take over the world.”

The premise of the paper, “Towards artificial general intelligence with hybrid Tianjic chip architecture,” is that to achieve AGI, computer chips need to change. That’s an idea supported by fervent activity these days in the land of computer chips, with lots of new chip designs being proposed specifically for machine learning.

The Tsinghua authors specifically propose that the mainstream machine learning of today needs to be merged in the same chip with what’s called “neuromorphic computing.” Neuromorphic computing, first conceived by Caltech professor Carver Mead in the early ’80s, has been an obsession for firms including IBM for years, with little practical result.

[Missing details about the chip] … For example, the part is said to have “reconfigurable” circuits, but how the circuits are to be reconfigured is never specified. It could be so-called “field programmable gate array,” or FPGA, technology or something else. Code for the project is not provided by the authors as it often is for such research; the authors offer to provide the code “on reasonable request.”

More important is the fact the chip may have a hard time stacking up to a lot of competing chips out there, says analyst Gwennap. …

What the paper calls ANN and SNN are two very different means of solving similar problems, kind of like rotating (helicopter) and fixed wing (airplane) are for aviation,” says Gwennap. “Ultimately, I expect ANN [?] and SNN [spiking neural network] to serve different end applications, but I don’t see a need to combine them in a single chip; you just end up with a chip that is OK for two things but not great for anything.”

But you also end up generating a lot of buzz, and given the tension between the U.S. and China over all things tech, and especially A.I., the notion China is stealing a march on the U.S. in artificial general intelligence — whatever that may be — is a summer sizzler of a headline.

ANN could be either artificial neural network or something mentioned earlier in Ray’s article, a shortened version of CANN [continuous attractor neural network].

Shelly Fan’s August 7, 2019 article for the SingularityHub is almost as enthusiastic about the work as the podcasters for Nature magazine  were (a little more about that later),

The study shows that China is readily nipping at the heels of Google, Facebook, NVIDIA, and other tech behemoths investing in developing new AI chip designs—hell, with billions in government investment it may have already had a head start. A sweeping AI plan from 2017 looks to catch up with the US on AI technology and application by 2020. By 2030, China’s aiming to be the global leader—and a champion for building general AI that matches humans in intellectual competence.

The country’s ambition is reflected in the team’s parting words.

“Our study is expected to stimulate AGI [artificial general intelligence] development by paving the way to more generalized hardware platforms,” said the authors, led by Dr. Luping Shi at Tsinghua University.

Using nanoscale fabrication, the team arranged 156 FCores, containing roughly 40,000 neurons and 10 million synapses, onto a chip less than a fifth of an inch in length and width. Initial tests showcased the chip’s versatility, in that it can run both SNNs and deep learning algorithms such as the popular convolutional neural network (CNNs) often used in machine vision.

Compared to IBM TrueNorth, the density of Tianjic’s cores increased by 20 percent, speeding up performance ten times and increasing bandwidth at least 100-fold, the team said. When pitted against GPUs, the current hardware darling of machine learning, the chip increased processing throughput up to 100 times, while using just a sliver (1/10,000) of energy.

BTW, Fan is a neuroscientist (from her SingularityHub profile page),

Shelly Xuelai Fan is a neuroscientist-turned-science writer. She completed her PhD in neuroscience at the University of British Columbia, where she developed novel treatments for neurodegeneration. While studying biological brains, she became fascinated with AI and all things biotech. Following graduation, she moved to UCSF [University of California at San Francisco] to study blood-based factors that rejuvenate aged brains. She is the co-founder of Vantastic Media, a media venture that explores science stories through text and video, and runs the award-winning blog NeuroFantastic.com. Her first book, “Will AI Replace Us?” (Thames & Hudson) will be out April 2019.

Onto Nature. Here’s a link to and a citation for the paper,

Towards artificial general intelligence with hybrid Tianjic chip architecture by Jing Pei, Lei Deng, Sen Song, Mingguo Zhao, Youhui Zhang, Shuang Wu, Guanrui Wang, Zhe Zou, Zhenzhi Wu, Wei He, Feng Chen, Ning Deng, Si Wu, Yu Wang, Yujie Wu, Zheyu Yang, Cheng Ma, Guoqi Li, Wentao Han, Huanglong Li, Huaqiang Wu, Rong Zhao, Yuan Xie & Luping Shi. Nature volume 572, pages106–111(2019) DOI: https//doi.org/10.1038/s41586-019-1424-8 Published: 31 July 2019 Issue Date: 01 August 2019

This paper is behind a paywall.

The July 31, 2019 Nature podcast, which includes a segment about the Tianjic chip research from China, which is at the 9 mins. 13 secs. mark (AI hardware) or you can scroll down about 55% of the way to the transcript of the interview with Luke Fleet, the Nature editor who dealt with the paper.

Some thoughts

The pundits put me in mind of my own reaction when I heard about phones that could take pictures. I didn’t see the point but, as it turned out, there was a perfectly good reason for combining what had been two separate activities into one device. It was no longer just a telephone and I had completely missed the point.

This too may be the case with the Tianjic chip. I think it’s too early to say whether or not it represents a new type of chip or if it’s a dead end.

A tangle of silver nanowires for brain-like action

I’ve been meaning to get to this news item from late 2019 as it features work from a team that I’ve been following for a number of years now. First mentioned here in an October 17, 2011 posting, James Gimzewski has been working with researchers at the University of California at Los Angeles (UCLA) and researchers at Japan’s National Institute for Materials Science (NIMS) on neuromorphic computing.

This particular research had a protracted rollout with the paper being published in October 2019 and the last news item about it being published in mid-December 2019.

A December 17, 2029 news item on Nanowerk was the first to alert me to this new work (Note: A link has been removed),

UCLA scientists James Gimzewski and Adam Stieg are part of an international research team that has taken a significant stride toward the goal of creating thinking machines.

Led by researchers at Japan’s National Institute for Materials Science, the team created an experimental device that exhibited characteristics analogous to certain behaviors of the brain — learning, memorization, forgetting, wakefulness and sleep. The paper, published in Scientific Reports (“Emergent dynamics of neuromorphic nanowire networks”), describes a network in a state of continuous flux.

A December 16, 2019 UCLA news release, which originated the news item, offers more detail (Note: A link has been removed),

“This is a system between order and chaos, on the edge of chaos,” said Gimzewski, a UCLA distinguished professor of chemistry and biochemistry, a member of the California NanoSystems Institute at UCLA and a co-author of the study. “The way that the device constantly evolves and shifts mimics the human brain. It can come up with different types of behavior patterns that don’t repeat themselves.”

The research is one early step along a path that could eventually lead to computers that physically and functionally resemble the brain — machines that may be capable of solving problems that contemporary computers struggle with, and that may require much less power than today’s computers do.

The device the researchers studied is made of a tangle of silver nanowires — with an average diameter of just 360 nanometers. (A nanometer is one-billionth of a meter.) The nanowires were coated in an insulating polymer about 1 nanometer thick. Overall, the device itself measured about 10 square millimeters — so small that it would take 25 of them to cover a dime.

Allowed to randomly self-assemble on a silicon wafer, the nanowires formed highly interconnected structures that are remarkably similar to those that form the neocortex, the part of the brain involved with higher functions such as language, perception and cognition.

One trait that differentiates the nanowire network from conventional electronic circuits is that electrons flowing through them cause the physical configuration of the network to change. In the study, electrical current caused silver atoms to migrate from within the polymer coating and form connections where two nanowires overlap. The system had about 10 million of these junctions, which are analogous to the synapses where brain cells connect and communicate.

The researchers attached two electrodes to the brain-like mesh to profile how the network performed. They observed “emergent behavior,” meaning that the network displayed characteristics as a whole that could not be attributed to the individual parts that make it up. This is another trait that makes the network resemble the brain and sets it apart from conventional computers.

After current flowed through the network, the connections between nanowires persisted for as much as one minute in some cases, which resembled the process of learning and memorization in the brain. Other times, the connections shut down abruptly after the charge ended, mimicking the brain’s process of forgetting.

In other experiments, the research team found that with less power flowing in, the device exhibited behavior that corresponds to what neuroscientists see when they use functional MRI scanning to take images of the brain of a sleeping person. With more power, the nanowire network’s behavior corresponded to that of the wakeful brain.

The paper is the latest in a series of publications examining nanowire networks as a brain-inspired system, an area of research that Gimzewski helped pioneer along with Stieg, a UCLA research scientist and an associate director of CNSI.

“Our approach may be useful for generating new types of hardware that are both energy-efficient and capable of processing complex datasets that challenge the limits of modern computers,” said Stieg, a co-author of the study.

The borderline-chaotic activity of the nanowire network resembles not only signaling within the brain but also other natural systems such as weather patterns. That could mean that, with further development, future versions of the device could help model such complex systems.

In other experiments, Gimzewski and Stieg already have coaxed a silver nanowire device to successfully predict statistical trends in Los Angeles traffic patterns based on previous years’ traffic data.

Because of their similarities to the inner workings of the brain, future devices based on nanowire technology could also demonstrate energy efficiency like the brain’s own processing. The human brain operates on power roughly equivalent to what’s used by a 20-watt incandescent bulb. By contrast, computer servers where work-intensive tasks take place — from training for machine learning to executing internet searches — can use the equivalent of many households’ worth of energy, with the attendant carbon footprint.

“In our studies, we have a broader mission than just reprogramming existing computers,” Gimzewski said. “Our vision is a system that will eventually be able to handle tasks that are closer to the way the human being operates.”

The study’s first author, Adrian Diaz-Alvarez, is from the International Center for Material Nanoarchitectonics at Japan’s National Institute for Materials Science. Co-authors include Tomonobu Nakayama and Rintaro Higuchi, also of NIMS; and Zdenka Kuncic at the University of Sydney in Australia.

Caption: (a) Micrograph of the neuromorphic network fabricated by this research team. The network contains of numerous junctions between nanowires, which operate as synaptic elements. When voltage is applied to the network (between the green probes), current pathways (orange) are formed in the network. (b) A Human brain and one of its neuronal networks. The brain is known to have a complex network structure and to operate by means of electrical signal propagation across the network. Credit: NIMS

A November 11, 2019 National Institute for Materials Science (Japan) press release (also on EurekAlert but dated December 25, 2019) first announced the news,

An international joint research team led by NIMS succeeded in fabricating a neuromorphic network composed of numerous metallic nanowires. Using this network, the team was able to generate electrical characteristics similar to those associated with higher order brain functions unique to humans, such as memorization, learning, forgetting, becoming alert and returning to calm. The team then clarified the mechanisms that induced these electrical characteristics.

The development of artificial intelligence (AI) techniques has been rapidly advancing in recent years and has begun impacting our lives in various ways. Although AI processes information in a manner similar to the human brain, the mechanisms by which human brains operate are still largely unknown. Fundamental brain components, such as neurons and the junctions between them (synapses), have been studied in detail. However, many questions concerning the brain as a collective whole need to be answered. For example, we still do not fully understand how the brain performs such functions as memorization, learning and forgetting, and how the brain becomes alert and returns to calm. In addition, live brains are difficult to manipulate in experimental research. For these reasons, the brain remains a “mysterious organ.” A different approach to brain research?in which materials and systems capable of performing brain-like functions are created and their mechanisms are investigated?may be effective in identifying new applications of brain-like information processing and advancing brain science.

The joint research team recently built a complex brain-like network by integrating numerous silver (Ag) nanowires coated with a polymer (PVP) insulating layer approximately 1 nanometer in thickness. A junction between two nanowires forms a variable resistive element (i.e., a synaptic element) that behaves like a neuronal synapse. This nanowire network, which contains a large number of intricately interacting synaptic elements, forms a “neuromorphic network”. When a voltage was applied to the neuromorphic network, it appeared to “struggle” to find optimal current pathways (i.e., the most electrically efficient pathways). The research team measured the processes of current pathway formation, retention and deactivation while electric current was flowing through the network and found that these processes always fluctuate as they progress, similar to the human brain’s memorization, learning, and forgetting processes. The observed temporal fluctuations also resemble the processes by which the brain becomes alert or returns to calm. Brain-like functions simulated by the neuromorphic network were found to occur as the huge number of synaptic elements in the network collectively work to optimize current transport, in the other words, as a result of self-organized and emerging dynamic processes..

The research team is currently developing a brain-like memory device using the neuromorphic network material. The team intends to design the memory device to operate using fundamentally different principles than those used in current computers. For example, while computers are currently designed to spend as much time and electricity as necessary in pursuit of absolutely optimum solutions, the new memory device is intended to make a quick decision within particular limits even though the solution generated may not be absolutely optimum. The team also hopes that this research will facilitate understanding of the brain’s information processing mechanisms.

This project was carried out by an international joint research team led by Tomonobu Nakayama (Deputy Director, International Center for Materials Nanoarchitectonics (WPI-MANA), NIMS), Adrian Diaz Alvarez (Postdoctoral Researcher, WPI-MANA, NIMS), Zdenka Kuncic (Professor, School of Physics, University of Sydney, Australia) and James K. Gimzewski (Professor, California NanoSystems Institute, University of California Los Angeles, USA).

Here at last is a link to and a citation for the paper,

Emergent dynamics of neuromorphic nanowire networks by Adrian Diaz-Alvarez, Rintaro Higuchi, Paula Sanz-Leon, Ido Marcus, Yoshitaka Shingaya, Adam Z. Stieg, James K. Gimzewski, Zdenka Kuncic & Tomonobu Nakayama. Scientific Reports volume 9, Article number: 14920 (2019) DOI: https://doi.org/10.1038/s41598-019-51330-6 Published: 17 October 2019

This paper is open access.

Of sleep, electric sheep, and thousands of artificial synapses on a chip

A close-up view of a new neuromorphic “brain-on-a-chip” that includes tens of thousands of memristors, or memory transistors. Credit: Peng Lin Courtesy: MIT

It’s hard to believe that a brain-on-a-chip might need sleep but that seems to be the case as far as the US Dept. of Energy’s Los Alamos National Laboratory is concerned. Before pursuing that line of thought, here’s some work from the Massachusetts Institute of Technology (MIT) involving memristors and a brain-on-a-chip. From a June 8, 2020 news item on ScienceDaily,

MIT engineers have designed a “brain-on-a-chip,” smaller than a piece of confetti, that is made from tens of thousands of artificial brain synapses known as memristors — silicon-based components that mimic the information-transmitting synapses in the human brain.

The researchers borrowed from principles of metallurgy to fabricate each memristor from alloys of silver and copper, along with silicon. When they ran the chip through several visual tasks, the chip was able to “remember” stored images and reproduce them many times over, in versions that were crisper and cleaner compared with existing memristor designs made with unalloyed elements.

Their results, published today in the journal Nature Nanotechnology, demonstrate a promising new memristor design for neuromorphic devices — electronics that are based on a new type of circuit that processes information in a way that mimics the brain’s neural architecture. Such brain-inspired circuits could be built into small, portable devices, and would carry out complex computational tasks that only today’s supercomputers can handle.

This ‘metallurgical’ approach differs somewhat from the protein nanowire approach used by the University of Massachusetts at Amherst team mentioned in my June 15, 2020 posting. Scientists are pursuing multiple pathways and we may find that we arrive with not ‘a single artificial brain but with many types of artificial brains.

A June 8, 2020 MIT news release (also on EurekAlert) provides more detail about this brain-on-a-chip,

“So far, artificial synapse networks exist as software. We’re trying to build real neural network hardware for portable artificial intelligence systems,” says Jeehwan Kim, associate professor of mechanical engineering at MIT. “Imagine connecting a neuromorphic device to a camera on your car, and having it recognize lights and objects and make a decision immediately, without having to connect to the internet. We hope to use energy-efficient memristors to do those tasks on-site, in real-time.”

Wandering ions

Memristors, or memory transistors [Note: Memristors are usually described as memory resistors; this is the first time I’ve seen ‘memory transistor’], are an essential element in neuromorphic computing. In a neuromorphic device, a memristor would serve as the transistor in a circuit, though its workings would more closely resemble a brain synapse — the junction between two neurons. The synapse receives signals from one neuron, in the form of ions, and sends a corresponding signal to the next neuron.

A transistor in a conventional circuit transmits information by switching between one of only two values, 0 and 1, and doing so only when the signal it receives, in the form of an electric current, is of a particular strength. In contrast, a memristor would work along a gradient, much like a synapse in the brain. The signal it produces would vary depending on the strength of the signal that it receives. This would enable a single memristor to have many values, and therefore carry out a far wider range of operations than binary transistors.

Like a brain synapse, a memristor would also be able to “remember” the value associated with a given current strength, and produce the exact same signal the next time it receives a similar current. This could ensure that the answer to a complex equation, or the visual classification of an object, is reliable — a feat that normally involves multiple transistors and capacitors.

Ultimately, scientists envision that memristors would require far less chip real estate than conventional transistors, enabling powerful, portable computing devices that do not rely on supercomputers, or even connections to the Internet.

Existing memristor designs, however, are limited in their performance. A single memristor is made of a positive and negative electrode, separated by a “switching medium,” or space between the electrodes. When a voltage is applied to one electrode, ions from that electrode flow through the medium, forming a “conduction channel” to the other electrode. The received ions make up the electrical signal that the memristor transmits through the circuit. The size of the ion channel (and the signal that the memristor ultimately produces) should be proportional to the strength of the stimulating voltage.

Kim says that existing memristor designs work pretty well in cases where voltage stimulates a large conduction channel, or a heavy flow of ions from one electrode to the other. But these designs are less reliable when memristors need to generate subtler signals, via thinner conduction channels.

The thinner a conduction channel, and the lighter the flow of ions from one electrode to the other, the harder it is for individual ions to stay together. Instead, they tend to wander from the group, disbanding within the medium. As a result, it’s difficult for the receiving electrode to reliably capture the same number of ions, and therefore transmit the same signal, when stimulated with a certain low range of current.

Borrowing from metallurgy

Kim and his colleagues found a way around this limitation by borrowing a technique from metallurgy, the science of melding metals into alloys and studying their combined properties.

“Traditionally, metallurgists try to add different atoms into a bulk matrix to strengthen materials, and we thought, why not tweak the atomic interactions in our memristor, and add some alloying element to control the movement of ions in our medium,” Kim says.

Engineers typically use silver as the material for a memristor’s positive electrode. Kim’s team looked through the literature to find an element that they could combine with silver to effectively hold silver ions together, while allowing them to flow quickly through to the other electrode.

The team landed on copper as the ideal alloying element, as it is able to bind both with silver, and with silicon.

“It acts as a sort of bridge, and stabilizes the silver-silicon interface,” Kim says.

To make memristors using their new alloy, the group first fabricated a negative electrode out of silicon, then made a positive electrode by depositing a slight amount of copper, followed by a layer of silver. They sandwiched the two electrodes around an amorphous silicon medium. In this way, they patterned a millimeter-square silicon chip with tens of thousands of memristors.

As a first test of the chip, they recreated a gray-scale image of the Captain America shield. They equated each pixel in the image to a corresponding memristor in the chip. They then modulated the conductance of each memristor that was relative in strength to the color in the corresponding pixel.

The chip produced the same crisp image of the shield, and was able to “remember” the image and reproduce it many times, compared with chips made of other materials.

The team also ran the chip through an image processing task, programming the memristors to alter an image, in this case of MIT’s Killian Court, in several specific ways, including sharpening and blurring the original image. Again, their design produced the reprogrammed images more reliably than existing memristor designs.

“We’re using artificial synapses to do real inference tests,” Kim says. “We would like to develop this technology further to have larger-scale arrays to do image recognition tasks. And some day, you might be able to carry around artificial brains to do these kinds of tasks, without connecting to supercomputers, the internet, or the cloud.”

Here’s a link to and a citation for the paper,

Alloying conducting channels for reliable neuromorphic computing by Hanwool Yeon, Peng Lin, Chanyeol Choi, Scott H. Tan, Yongmo Park, Doyoon Lee, Jaeyong Lee, Feng Xu, Bin Gao, Huaqiang Wu, He Qian, Yifan Nie, Seyoung Kim & Jeehwan Kim. Nature Nanotechnology (2020 DOI: https://doi.org/10.1038/s41565-020-0694-5 Published: 08 June 2020

This paper is behind a paywall.

Electric sheep and sleeping androids

I find it impossible to mention that androids might need sleep without reference to Philip K. Dick’s 1968 novel, “Do Androids Dream of Electric Sheep?”; its Wikipedia entry is here.

June 8, 2020 Intelligent machines of the future may need to sleep as much as we do. Intelligent machines of the future may need to sleep as much as we do. Courtesy: Los Alamos National Laboratory

As it happens, I’m not the only one who felt the need to reference the novel, from a June 8, 2020 news item on ScienceDaily,

No one can say whether androids will dream of electric sheep, but they will almost certainly need periods of rest that offer benefits similar to those that sleep provides to living brains, according to new research from Los Alamos National Laboratory.

“We study spiking neural networks, which are systems that learn much as living brains do,” said Los Alamos National Laboratory computer scientist Yijing Watkins. “We were fascinated by the prospect of training a neuromorphic processor in a manner analogous to how humans and other biological systems learn from their environment during childhood development.”

Watkins and her research team found that the network simulations became unstable after continuous periods of unsupervised learning. When they exposed the networks to states that are analogous to the waves that living brains experience during sleep, stability was restored. “It was as though we were giving the neural networks the equivalent of a good night’s rest,” said Watkins.

A June 8, 2020 Los Alamos National Laboratory (LANL) news release (also on EurekAlert), which originated the news item, describes the research team’s presentation,

The discovery came about as the research team worked to develop neural networks that closely approximate how humans and other biological systems learn to see. The group initially struggled with stabilizing simulated neural networks undergoing unsupervised dictionary training, which involves classifying objects without having prior examples to compare them to.

“The issue of how to keep learning systems from becoming unstable really only arises when attempting to utilize biologically realistic, spiking neuromorphic processors or when trying to understand biology itself,” said Los Alamos computer scientist and study coauthor Garrett Kenyon. “The vast majority of machine learning, deep learning, and AI researchers never encounter this issue because in the very artificial systems they study they have the luxury of performing global mathematical operations that have the effect of regulating the overall dynamical gain of the system.”

The researchers characterize the decision to expose the networks to an artificial analog of sleep as nearly a last ditch effort to stabilize them. They experimented with various types of noise, roughly comparable to the static you might encounter between stations while tuning a radio. The best results came when they used waves of so-called Gaussian noise, which includes a wide range of frequencies and amplitudes. They hypothesize that the noise mimics the input received by biological neurons during slow-wave sleep. The results suggest that slow-wave sleep may act, in part, to ensure that cortical neurons maintain their stability and do not hallucinate.

The groups’ next goal is to implement their algorithm on Intel’s Loihi neuromorphic chip. They hope allowing Loihi to sleep from time to time will enable it to stably process information from a silicon retina camera in real time. If the findings confirm the need for sleep in artificial brains, we can probably expect the same to be true of androids and other intelligent machines that may come about in the future.

Watkins will be presenting the research at the Women in Computer Vision Workshop on June 14 [2020] in Seattle.

The 2020 Women in Computer Vition Workshop (WICV) website is here. As is becoming standard practice for these times, the workshop was held in a virtual environment. Here’s a link to and a citation for the poster presentation paper,

Using Sinusoidally-Modulated Noise as a Surrogate for Slow-Wave Sleep to
Accomplish Stable Unsupervised Dictionary Learning in a Spike-Based Sparse Coding Model
by Yijing Watkins, Edward Kim, Andrew Sornborger and Garrett T. Kenyon. Women in Computer Vision Workshop on June 14, 2020 in Seattle, Washington (state)

This paper is open access for now.

Neuromorphic computing with voltage usage comparable to human brains

Part of neuromorphic computing’s appeal is the promise of using less energy because, as it turns out, the human brain uses small amounts of energy very efficiently. A team of researchers at the University of Massachusetts at Amherst have developed function in the same range of voltages as the human brain. From an April 20, 2020 news item on ScienceDaily,

Only 10 years ago, scientists working on what they hoped would open a new frontier of neuromorphic computing could only dream of a device using miniature tools called memristors that would function/operate like real brain synapses.

But now a team at the University of Massachusetts Amherst has discovered, while on their way to better understanding protein nanowires, how to use these biological, electricity conducting filaments to make a neuromorphic memristor, or “memory transistor,” device. It runs extremely efficiently on very low power, as brains do, to carry signals between neurons. Details are in Nature Communications.

An April 20, 2020 University of Massachusetts at Amherst news release (also on EurekAlert), which originated the news items, dives into detail about how these researchers were able to achieve bio-voltages,

As first author Tianda Fu, a Ph.D. candidate in electrical and computer engineering, explains, one of the biggest hurdles to neuromorphic computing, and one that made it seem unreachable, is that most conventional computers operate at over 1 volt, while the brain sends signals called action potentials between neurons at around 80 millivolts – many times lower. Today, a decade after early experiments, memristor voltage has been achieved in the range similar to conventional computer, but getting below that seemed improbable, he adds.

Fu reports that using protein nanowires developed at UMass Amherst from the bacterium Geobacter by microbiologist and co-author Derek Lovely, he has now conducted experiments where memristors have reached neurological voltages. Those tests were carried out in the lab of electrical and computer engineering researcher and co-author Jun Yao.

Yao says, “This is the first time that a device can function at the same voltage level as the brain. People probably didn’t even dare to hope that we could create a device that is as power-efficient as the biological counterparts in a brain, but now we have realistic evidence of ultra-low power computing capabilities. It’s a concept breakthrough and we think it’s going to cause a lot of exploration in electronics that work in the biological voltage regime.”

Lovely points out that Geobacter’s electrically conductive protein nanowires offer many advantages over expensive silicon nanowires, which require toxic chemicals and high-energy processes to produce. Protein nanowires also are more stable in water or bodily fluids, an important feature for biomedical applications. For this work, the researchers shear nanowires off the bacteria so only the conductive protein is used, he adds.

Fu says that he and Yao had set out to put the purified nanowires through their paces, to see what they are capable of at different voltages, for example. They experimented with a pulsing on-off pattern of positive-negative charge sent through a tiny metal thread in a memristor, which creates an electrical switch.

They used a metal thread because protein nanowires facilitate metal reduction, changing metal ion reactivity and electron transfer properties. Lovely says this microbial ability is not surprising, because wild bacterial nanowires breathe and chemically reduce metals to get their energy the way we breathe oxygen.

As the on-off pulses create changes in the metal filaments, new branching and connections are created in the tiny device, which is 100 times smaller than the diameter of a human hair, Yao explains. It creates an effect similar to learning – new connections – in a real brain. He adds, “You can modulate the conductivity, or the plasticity of the nanowire-memristor synapse so it can emulate biological components for brain-inspired computing. Compared to a conventional computer, this device has a learning capability that is not software-based.”

Fu recalls, “In the first experiments we did, the nanowire performance was not satisfying, but it was enough for us to keep going.” Over two years, he saw improvement until one fateful day when his and Yao’s eyes were riveted by voltage measurements appearing on a computer screen.

“I remember the day we saw this great performance. We watched the computer as current voltage sweep was being measured. It kept doing down and down and we said to each other, ‘Wow, it’s working.’ It was very surprising and very encouraging.”

Fu, Yao, Lovely and colleagues plan to follow up this discovery with more research on mechanisms, and to “fully explore the chemistry, biology and electronics” of protein nanowires in memristors, Fu says, plus possible applications, which might include a device to monitor heart rate, for example. Yao adds, “This offers hope in the feasibility that one day this device can talk to actual neurons in biological systems.”

That last comment has me wondering about why you would want to have your device talk to actual neurons. For neuroprosthetics perhaps?

Here’s a link to and a citation for the paper,

Bioinspired bio-voltage memristors by Tianda Fu, Xiaomeng Liu, Hongyan Gao, Joy E. Ward, Xiaorong Liu, Bing Yin, Zhongrui Wang, Ye Zhuo, David J. F. Walker, J. Joshua Yang, Jianhan Chen, Derek R. Lovley & Jun Yao. Nature Communications volume 11, Article number: 1861 (2020) DOI: https://doi.org/10.1038/s41467-020-15759-y Published: 20 April 2020

This paper is open access.

There is an illustration of the work

Caption: A graphic depiction of protein nanowires (green) harvested from microbe Geobacter (orange) facilitate the electronic memristor device (silver) to function with biological voltages, emulating the neuronal components (blue junctions) in a brain. Credit: UMass Amherst/Yao lab

New design directions to increase variety, efficiency, selectivity and reliability for memristive devices

A May 11, 2020 news item on ScienceDaily provides a description of the current ‘memristor scene’ along with an announcement about a piece of recent research,

Scientists around the world are intensively working on memristive devices, which are capable in extremely low power operation and behave similarly to neurons in the brain. Researchers from the Jülich Aachen Research Alliance (JARA) and the German technology group Heraeus have now discovered how to systematically control the functional behaviour of these elements. The smallest differences in material composition are found crucial: differences so small that until now experts had failed to notice them. The researchers’ design directions could help to increase variety, efficiency, selectivity and reliability for memristive technology-based applications, for example for energy-efficient, non-volatile storage devices or neuro-inspired computers.

Memristors are considered a highly promising alternative to conventional nanoelectronic elements in computer Chips [sic]. Because of the advantageous functionalities, their development is being eagerly pursued by many companies and research institutions around the world. The Japanese corporation NEC installed already the first prototypes in space satellites back in 2017. Many other leading companies such as Hewlett Packard, Intel, IBM, and Samsung are working to bring innovative types of computer and storage devices based on memristive elements to market.

Fundamentally, memristors are simply “resistors with memory,” in which high resistance can be switched to low resistance and back again. This means in principle that the devices are adaptive, similar to a synapse in a biological nervous system. “Memristive elements are considered ideal candidates for neuro-inspired computers modelled on the brain, which are attracting a great deal of interest in connection with deep learning and artificial intelligence,” says Dr. Ilia Valov of the Peter Grünberg Institute (PGI-7) at Forschungszentrum Jülich.

In the latest issue of the open access journal Science Advances, he and his team describe how the switching and neuromorphic behaviour of memristive elements can be selectively controlled. According to their findings, the crucial factor is the purity of the switching oxide layer. “Depending on whether you use a material that is 99.999999 % pure, and whether you introduce one foreign atom into ten million atoms of pure material or into one hundred atoms, the properties of the memristive elements vary substantially” says Valov.

A May 11, 2020 Forschungszentrum Juelich press release (also on EurekAlert), which originated the news item, delves into the theme of increasing control over memristive systems,

This effect had so far been overlooked by experts. It can be used very specifically for designing memristive systems, in a similar way to doping semiconductors in information technology. “The introduction of foreign atoms allows us to control the solubility and transport properties of the thin oxide layers,” explains Dr. Christian Neumann of the technology group Heraeus. He has been contributing his materials expertise to the project ever since the initial idea was conceived in 2015.

“In recent years there has been remarkable progress in the development and use of memristive devices, however that progress has often been achieved on a purely empirical basis,” according to Valov. Using the insights that his team has gained, manufacturers could now methodically develop memristive elements selecting the functions they need. The higher the doping concentration, the slower the resistance of the elements changes as the number of incoming voltage pulses increases and decreases, and the more stable the resistance remains. “This means that we have found a way for designing types of artificial synapses with differing excitability,” explains Valov.

Design specification for artificial synapses

The brain’s ability to learn and retain information can largely be attributed to the fact that the connections between neurons are strengthened when they are frequently used. Memristive devices, of which there are different types such as electrochemical metallization cells (ECMs) or valence change memory cells (VCMs), behave similarly. When these components are used, the conductivity increases as the number of incoming voltage pulses increases. The changes can also be reversed by applying voltage pulses of the opposite polarity.

The JARA researchers conducted their systematic experiments on ECMs, which consist of a copper electrode, a platinum electrode, and a layer of silicon dioxide between them. Thanks to the cooperation with Heraeus researchers, the JARA scientists had access to different types of silicon dioxide: one with a purity of 99.999999 % – also called 8N silicon dioxide – and others containing 100 to 10,000 ppm (parts per million) of foreign atoms. The precisely doped glass used in their experiments was specially developed and manufactured by quartz glass specialist Heraeus Conamic, which also holds the patent for the procedure. Copper and protons acted as mobile doping agents, while aluminium and gallium were used as non-volatile doping.

Synapses, the connections between neurons, have the ability to transmit signals with varying degrees of strength when they are excited by a quick succession of electrical impulses. One effect of this repeated activity is to increase the concentration of calcium ions, with the result that more neurotransmitters are emitted. Depending on the activity, other effects cause long-term structural changes, which impact the strength of the transmission for several hours, or potentially even for the rest of the person’s life. Memristive elements allow the strength of the electrical transmission to be changed in a similar way to synaptic connections, by applying a voltage. In electrochemical metallization cells (ECMs), a metallic filament develops between the two metal electrodes, thus increasing conductivity. Applying voltage pulses with reversed polarity causes the filament to shrink again until the cell reaches its initial high resistance state. Copyright: Forschungszentrum Jülich / Tobias Schlößer

Record switching time confirms theory

Based on their series of experiments, the researchers were able to show that the ECMs’ switching times change as the amount of doping atoms changes. If the switching layer is made of 8N silicon dioxide, the memristive component switches in only 1.4 nanoseconds. To date, the fastest value ever measured for ECMs had been around 10 nanoseconds. By doping the oxide layer of the components with up to 10,000 ppm of foreign atoms, the switching time was prolonged into the range of milliseconds. “We can also theoretically explain our results. This is helping us to understand the physico-chemical processes on the nanoscale and apply this knowledge in the practice” says Valov. Based on generally applicable theoretical considerations, supported by experimental results, some also documented in the literature, he is convinced that the doping/impurity effect occurs and can be employed in all types memristive elements.

Top: In memristive elements (ECMs) with an undoped, high-purity switching layer of silicon oxide (SiO2), copper ions can move very fast. A filament of copper atoms forms correspondingly fast on the platinum electrode. This increases the total device conductivity respectively the capacity. Due to the high mobility of the ions, however, this filament is unstable at low forming voltages. Center: Gallium ions (Ga3+), which are introduced into the cell (non-volatile doping), bind copper ions (Cu2+) in the switching layer. The movement of the ions slows down, leading to lower switching times, but the filament, once formed remains longer stable. Bottom: Doping with aluminium ions (Al3+) slows down the process even more, since aluminium ions bind copper ions even stronger than gallium ions. Filament growth is even slower, while at the same time the stability of the filament is further increased. Depending on the chemical properties of the introduced doping elements, memristive cells – the artificial synapses – can be created with tailor-made switching and neuromorphic properties. Copyright: Forschungszentrum Jülich / Tobias Schloesser

Here’s a link to and a citation for the paper,

Design of defect-chemical properties and device performance in memristive systems by M. Lübben, F. Cüppers, J. Mohr, M. von Witzleben, U. Breuer, R. Waser, C. Neumann, and I. Valov. Science Advances 08 May 2020: Vol. 6, no. 19, eaaz9079 DOI: 10.1126/sciadv.aaz9079

This paper is open access.

For anyone curious about the German technology group, Heraeus, there’s a fascinating history in its Wikipedia entry. The technology company was formally founded in 1851 but it can be traced back to the 17th century and the founding family’s apothecary.

Connecting biological and artificial neurons (in UK, Switzerland, & Italy) over the web

Caption: The virtual lab connecting Southampton, Zurich and Padova. Credit: University of Southampton

A February 26, 2020 University of Southampton press release (also on EurekAlert) describes this work,

Research on novel nanoelectronics devices led by the University of Southampton enabled brain neurons and artificial neurons to communicate with each other. This study has for the first time shown how three key emerging technologies can work together: brain-computer interfaces, artificial neural networks and advanced memory technologies (also known as memristors). The discovery opens the door to further significant developments in neural and artificial intelligence research.

Brain functions are made possible by circuits of spiking neurons, connected together by microscopic, but highly complex links called ‘synapses’. In this new study, published in the scientific journal Nature Scientific Reports, the scientists created a hybrid neural network where biological and artificial neurons in different parts of the world were able to communicate with each other over the internet through a hub of artificial synapses made using cutting-edge nanotechnology. This is the first time the three components have come together in a unified network.

During the study, researchers based at the University of Padova in Italy cultivated rat neurons in their laboratory, whilst partners from the University of Zurich and ETH Zurich created artificial neurons on Silicon microchips. The virtual laboratory was brought together via an elaborate setup controlling nanoelectronic synapses developed at the University of Southampton. These synaptic devices are known as memristors.

The Southampton based researchers captured spiking events being sent over the internet from the biological neurons in Italy and then distributed them to the memristive synapses. Responses were then sent onward to the artificial neurons in Zurich also in the form of spiking activity. The process simultaneously works in reverse too; from Zurich to Padova. Thus, artificial and biological neurons were able to communicate bidirectionally and in real time.

Themis Prodromakis, Professor of Nanotechnology and Director of the Centre for Electronics Frontiers at the University of Southampton said “One of the biggest challenges in conducting research of this kind and at this level has been integrating such distinct cutting edge technologies and specialist expertise that are not typically found under one roof. By creating a virtual lab we have been able to achieve this.”

The researchers now anticipate that their approach will ignite interest from a range of scientific disciplines and accelerate the pace of innovation and scientific advancement in the field of neural interfaces research. In particular, the ability to seamlessly connect disparate technologies across the globe is a step towards the democratisation of these technologies, removing a significant barrier to collaboration.

Professor Prodromakis added “We are very excited with this new development. On one side it sets the basis for a novel scenario that was never encountered during natural evolution, where biological and artificial neurons are linked together and communicate across global networks; laying the foundations for the Internet of Neuro-electronics. On the other hand, it brings new prospects to neuroprosthetic technologies, paving the way towards research into replacing dysfunctional parts of the brain with AI [artificial intelligence] chips.”

I’m fascinated by this work and after taking a look at the paper, I have to say, the paper is surprisingly accessible. In other words, I think I get the general picture. For example (from the Introduction to the paper; citation and link follow further down),

… To emulate plasticity, the memristor MR1 is operated as a two-terminal device through a control system that receives pre- and post-synaptic depolarisations from one silicon neuron (ANpre) and one biological neuron (BN), respectively. …

If I understand this properly, they’ve integrated a biological neuron and an artificial neuron in a single system across three countries.

For those who care to venture forth, here’s a link and a citation for the paper,

Memristive synapses connect brain and silicon spiking neurons by Alexantrou Serb, Andrea Corna, Richard George, Ali Khiat, Federico Rocchi, Marco Reato, Marta Maschietto, Christian Mayr, Giacomo Indiveri, Stefano Vassanelli & Themistoklis Prodromakis. Scientific Reports volume 10, Article number: 2590 (2020) DOI: https://doi.org/10.1038/s41598-020-58831-9 Published 25 February 2020

The paper is open access.

A lipid-based memcapacitor,for neuromorphic computing

Caption: Researchers at ORNL’s Center for Nanophase Materials Sciences demonstrated the first example of capacitance in a lipid-based biomimetic membrane, opening nondigital routes to advanced, brain-like computation. Credit: Michelle Lehman/Oak Ridge National Laboratory, U.S. Dept. of Energy

The last time I wrote about memcapacitors (June 30, 2014 posting: Memristors, memcapacitors, and meminductors for faster computers), the ideas were largely theoretical; I believe this work is the first research I’ve seen on the topic. From an October 17, 2019 news item on ScienceDaily,

Researchers at the Department of Energy’s Oak Ridge National Laboratory ]ORNL], the University of Tennessee and Texas A&M University demonstrated bio-inspired devices that accelerate routes to neuromorphic, or brain-like, computing.

Results published in Nature Communications report the first example of a lipid-based “memcapacitor,” a charge storage component with memory that processes information much like synapses do in the brain. Their discovery could support the emergence of computing networks modeled on biology for a sensory approach to machine learning.

An October 16, 2019 ORNL news release (also on EurekAlert but published Oct. 17, 2019), which originated the news item, provides more detail about the work,

“Our goal is to develop materials and computing elements that work like biological synapses and neurons—with vast interconnectivity and flexibility—to enable autonomous systems that operate differently than current computing devices and offer new functionality and learning capabilities,” said Joseph Najem, a recent postdoctoral researcher at ORNL’s Center for Nanophase Materials Sciences, a DOE Office of Science User Facility, and current assistant professor of mechanical engineering at Penn State.

The novel approach uses soft materials to mimic biomembranes and simulate the way nerve cells communicate with one another.

The team designed an artificial cell membrane, formed at the interface of two lipid-coated water droplets in oil, to explore the material’s dynamic, electrophysiological properties. At applied voltages, charges build up on both sides of the membrane as stored energy, analogous to the way capacitors work in traditional electric circuits.

But unlike regular capacitors, the memcapacitor can “remember” a previously applied voltage and—literally—shape how information is processed. The synthetic membranes change surface area and thickness depending on electrical activity. These shapeshifting membranes could be tuned as adaptive filters for specific biophysical and biochemical signals.

“The novel functionality opens avenues for nondigital signal processing and machine learning modeled on nature,” said ORNL’s Pat Collier, a CNMS staff research scientist.

A distinct feature of all digital computers is the separation of processing and memory. Information is transferred back and forth from the hard drive and the central processor, creating an inherent bottleneck in the architecture no matter how small or fast the hardware can be.

Neuromorphic computing, modeled on the nervous system, employs architectures that are fundamentally different in that memory and signal processing are co-located in memory elements—memristors, memcapacitors and meminductors.

These “memelements” make up the synaptic hardware of systems that mimic natural information processing, learning and memory.

Systems designed with memelements offer advantages in scalability and low power consumption, but the real goal is to carve out an alternative path to artificial intelligence, said Collier.

Tapping into biology could enable new computing possibilities, especially in the area of “edge computing,” such as wearable and embedded technologies that are not connected to a cloud but instead make on-the-fly decisions based on sensory input and past experience.

Biological sensing has evolved over billions of years into a highly sensitive system with receptors in cell membranes that are able to pick out a single molecule of a specific odor or taste. “This is not something we can match digitally,” Collier said.

Digital computation is built around digital information, the binary language of ones and zeros coursing through electronic circuits. It can emulate the human brain, but its solid-state components do not compute sensory data the way a brain does.

“The brain computes sensory information pushed through synapses in a neural network that is reconfigurable and shaped by learning,” said Collier. “Incorporating biology—using biomembranes that sense bioelectrochemical information—is key to developing the functionality of neuromorphic computing.”

While numerous solid-state versions of memelements have been demonstrated, the team’s biomimetic elements represent new opportunities for potential “spiking” neural networks that can compute natural data in natural ways.

Spiking neural networks are intended to simulate the way neurons spike with electrical potential and, if the signal is strong enough, pass it on to their neighbors through synapses, carving out learning pathways that are pruned over time for efficiency.

A bio-inspired version with analog data processing is a distant aim. Current early-stage research focuses on developing the components of bio-circuitry.

“We started with the basics, a memristor that can weigh information via conductance to determine if a spike is strong enough to be broadcast through a network of synapses connecting neurons,” said Collier. “Our memcapacitor goes further in that it can actually store energy as an electric charge in the membrane, enabling the complex ‘integrate and fire’ activity of neurons needed to achieve dense networks capable of brain-like computation.”

The team’s next steps are to explore new biomaterials and study simple networks to achieve more complex brain-like functionalities with memelements.

Here’s a link to and a citation for the paper,

Dynamical nonlinear memory capacitance in biomimetic membranes by Joseph S. Najem, Md Sakib Hasan, R. Stanley Williams, Ryan J. Weiss, Garrett S. Rose, Graham J. Taylor, Stephen A. Sarles & C. Patrick Collier. Nature Communications volume 10, Article number: 3239 (2019) DOI: DOIhttps://doi.org/10.1038/s41467-019-11223-8 Published July 19, 2019

This paper is open access.

One final comment, you might recognize one of the authors (R. Stanley Williams) who in 2008 helped launch ‘memristor’ research.

Second order memristor

I think this is my first encounter with a second-order memristor. An August 28, 2019 news item on Nanowerk announces the research (Note: A link has been removed),

Researchers from the Moscow Institute of Physics and Technology [MIPT} have created a device that acts like a synapse in the living brain, storing information and gradually forgetting it when not accessed for a long time. Known as a second-order memristor, the new device is based on hafnium oxide and offers prospects for designing analog neurocomputers imitating the way a biological brain learns.

An August 28, 2019 MIPT press release (also on EurekAlert), which originated the news item, provides an explanation for neuromorphic computing (analog neurocomputers; brainlike computing), the difference between a first-order and second-order memristor, and an in depth view of the research,

Neurocomputers, which enable artificial intelligence, emulate the way the brain works. It stores data in the form of synapses, a network of connections between the nerve cells, or neurons. Most neurocomputers have a conventional digital architecture and use mathematical models to invoke virtual neurons and synapses.

Alternatively, an actual on-chip electronic component could stand for each neuron and synapse in the network. This so-called analog approach has the potential to drastically speed up computations and reduce energy costs.

The core component of a hypothetical analog neurocomputer is the memristor. The word is a portmanteau of “memory” and “resistor,” which pretty much sums up what it is: a memory cell acting as a resistor. Loosely speaking, a high resistance encodes a zero, and a low resistance encodes a one. This is analogous to how a synapse conducts a signal between two neurons (one), while the absence of a synapse results in no signal, a zero.

But there is a catch: In an actual brain, the active synapses tend to strengthen over time, while the opposite is true for inactive ones. This phenomenon known as synaptic plasticity is one of the foundations of natural learning and memory. It explains the biology of cramming for an exam and why our seldom accessed memories fade.

Proposed in 2015, the second-order memristor is an attempt to reproduce natural memory, complete with synaptic plasticity. The first mechanism for implementing this involves forming nanosized conductive bridges across the memristor. While initially decreasing resistance, they naturally decay with time, emulating forgetfulness.

“The problem with this solution is that the device tends to change its behavior over time and breaks down after prolonged operation,” said the study’s lead author Anastasia Chouprik from MIPT’s Neurocomputing Systems Lab. “The mechanism we used to implement synaptic plasticity is more robust. In fact, after switching the state of the system 100 billion times, it was still operating normally, so my colleagues stopped the endurance test.”

Instead of nanobridges, the MIPT team relied on hafnium oxide to imitate natural memory. This material is ferroelectric: Its internal bound charge distribution — electric polarization — changes in response to an external electric field. If the field is then removed, the material retains its acquired polarization, the way a ferromagnet remains magnetized.

The physicists implemented their second-order memristor as a ferroelectric tunnel junction — two electrodes interlaid with a thin hafnium oxide film (fig. 1, right). The device can be switched between its low and high resistance states by means of electric pulses, which change the ferroelectric film’s polarization and thus its resistance.

“The main challenge that we faced was figuring out the right ferroelectric layer thickness,” Chouprik added. “Four nanometers proved to be ideal. Make it just one nanometer thinner, and the ferroelectric properties are gone, while a thicker film is too wide a barrier for the electrons to tunnel through. And it is only the tunneling current that we can modulate by switching polarization.”

What gives hafnium oxide an edge over other ferroelectric materials, such as barium titanate, is that it is already used by current silicon technology. For example, Intel has been manufacturing microchips based on a hafnium compound since 2007. This makes introducing hafnium-based devices like the memristor reported in this story far easier and cheaper than those using a brand-new material.

In a feat of ingenuity, the researchers implemented “forgetfulness” by leveraging the defects at the interface between silicon and hafnium oxide. Those very imperfections used to be seen as a detriment to hafnium-based microprocessors, and engineers had to find a way around them by incorporating other elements into the compound. Instead, the MIPT team exploited the defects, which make memristor conductivity die down with time, just like natural memories.

Vitalii Mikheev, the first author of the paper, shared the team’s future plans: “We are going to look into the interplay between the various mechanisms switching the resistance in our memristor. It turns out that the ferroelectric effect may not be the only one involved. To further improve the devices, we will need to distinguish between the mechanisms and learn to combine them.”

According to the physicists, they will move on with the fundamental research on the properties of hafnium oxide to make the nonvolatile random access memory cells more reliable. The team is also investigating the possibility of transferring their devices onto a flexible substrate, for use in flexible electronics.

Last year, the researchers offered a detailed description of how applying an electric field to hafnium oxide films affects their polarization. It is this very process that enables reducing ferroelectric memristor resistance, which emulates synapse strengthening in a biological brain. The team also works on neuromorphic computing systems with a digital architecture.

MIPT has provided this image illustrating the research,

Caption: The left image shows a synapse from a biological brain, the inspiration behind its artificial analogue (right). The latter is a memristor device implemented as a ferroelectric tunnel junction — that is, a thin hafnium oxide film (pink) interlaid between a titanium nitride electrode (blue cable) and a silicon substrate (marine blue), which doubles up as the second electrode. Electric pulses switch the memristor between its high and low resistance states by changing hafnium oxide polarization, and therefore its conductivity. Credit: Elena Khavina/MIPT Press Office

Here’s a link to and a citation for the paper,

Ferroelectric Second-Order Memristor by Vitalii Mikheev, Anastasia Chouprik, Yury Lebedinskii, Sergei Zarubin, Yury Matveyev, Ekaterina Kondratyuk, Maxim G. Kozodaev, Andrey M. Markeev, Andrei Zenkevich, Dmitrii Negrov. ACS Appl. Mater. Interfaces 2019113532108-32114 DOI: https://doi.org/10.1021/acsami.9b08189 Publication Date:August 12, 2019 Copyright © 2019 American Chemical Society

This paper is behind a paywall.

Memristor-based neural network and the biosimilar principle of learning

Once you get past the technical language (there’s a lot of it), you’ll find that they make the link between biomimicry and memristors explicit. Admittedly I’m not an expert but if I understand the research correctly, the scientists are suggesting that the algorithms used in machine learning today cannot allow memristors to be properly integrated for use in true neuromorphic computing and this work from Russia and Greece points to a new paradigm. If you understand it differently, please do let me know in the comments.

A July 12, 2019 news item on Nanowerk kicks things off (Note: A link has been removed),

Lobachevsky University scientists together with their colleagues from the National Research Center “Kurchatov Institute” (Moscow) and the National Research Center “Demokritos” (Athens) are working on the hardware implementation of a spiking neural network based on memristors.

The key elements of such a network, along with pulsed neurons, are artificial synaptic connections that can change the strength (weight) of connection between neurons during the learning (Microelectronic Engineering, “Yttria-stabilized zirconia cross-point memristive devices for neuromorphic applications”).

For this purpose, memristive devices based on metal-oxide-metal nanostructures developed at the UNN Physics and Technology Research Institute (PTRI) are suitable, but their use in specific spiking neural network architectures developed at the Kurchatov Institute requires demonstration of biologically plausible learning principles.

Caption: Cross-section image of the metal-oxide-metal memristive structure based on ZrO2(Y) polycrystalline film (a); corresponding schematic view of the cross-point memristive device (b); STDP dependencies of memristive device conductance changes for different delay values between pre- and postsynaptic neuron spikes (c); photographs of a microchip and an array of memristive devices in a standard cermet casing (d); the simplest spiking neural network architecture learning on the basis of local rules for changing memristive weights (e). Credit: Lobachevsky University

A July 12, 2019 (?) Lobachevsky University press release (also on EurekAlert), which originated the news item, delves further into the work,

The biological mechanism of learning of neural systems is described by Hebb’s rule, according to which learning occurs as a result of an increase in the strength of connection  (synaptic weight) between simultaneously active neurons, which indicates the presence of a causal relationship in their excitation. One of the clarifying forms of this fundamental rule is plasticity, which depends on the time of arrival of pulses (Spike-Timing Dependent Plasticity – STDP).

In accordance with STDP, synaptic weight increases if the postsynaptic neuron generates a pulse (spike) immediately after the presynaptic one, and vice versa, the synaptic weight decreases if the postsynaptic neuron generates a spike right before the presynaptic one. Moreover, the smaller the time difference Δt between the pre- and postsynaptic spikes, the more pronounced the weight change will be.

According to one of the researchers, Head of the UNN PTRI laboratory Alexei Mikhailov, in order to demonstrate the STDP principle, memristive nanostructures based on yttria-stabilized zirconia (YSZ) thin films were used. YSZ is a well-known solid-state electrolyte with high oxygen ion mobility.

“Due to a specified concentration of oxygen vacancies, which is determined by the controlled concentration of yttrium impurities, and the heterogeneous structure of the films obtained by magnetron sputtering, such memristive structures demonstrate controlled bipolar switching between different resistive states in a wide resistance range. The switching is associated with the formation and destruction of conductive channels along grain boundaries in the polycrystalline ZrO2 (Y) film,” notes Alexei Mikhailov.

An array of memristive devices for research was implemented in the form of a microchip mounted in a standard cermet casing, which facilitates the integration of the array into a neural network’s analog circuit. The full technological cycle for creating memristive microchips is currently implemented at the UNN PTRI. In the future, it is possible to scale the devices down to the minimum size of about 50 nm, as was established by Greek partners.
Our studies of the dynamic plasticity of the memoristive devices, continues Alexey Mikhailov, have shown that the form of the conductance change depending on Δt is in good agreement with the STDP learning rules. It should be also noted that if the initial value of the memristor conductance is close to the maximum, it is easy to reduce the corresponding weight while it is difficult to enhance it, and in the case of a memristor with a minimum conductance in the initial state, it is difficult to reduce its weight, but it is easy to enhance it.

According to Vyacheslav Demin, director-coordinator in the area of nature-like technologies of the Kurchatov Institute, who is one of the ideologues of this work, the established pattern of change in the memristor conductance clearly demonstrates the possibility of hardware implementation of the so-called local learning rules. Such rules for changing the strength of synaptic connections depend only on the values ​​of variables that are present locally at each time point (neuron activities and current weights).

“This essentially distinguishes such principle from the traditional learning algorithm, which is based on global rules for changing weights, using information on the error values ​​at the current time point for each neuron of the output neural network layer (in a widely popular group of error back propagation methods). The traditional principle is not biosimilar, it requires “external” (expert) knowledge of the correct answers for each example presented to the network (that is, they do not have the property of self-learning). This principle is difficult to implement on the basis of memristors, since it requires controlled precise changes of memristor conductances, as opposed to local rules. Such precise control is not always possible due to the natural variability (a wide range of parameters) of memristors as analog elements,” says Vyacheslav Demin.

Local learning rules of the STDP type implemented in hardware on memristors provide the basis for autonomous (“unsupervised”) learning of a spiking neural network. In this case, the final state of the network does not depend on its initial state, but depends only on the learning conditions (a specific sequence of pulses). According to Vyacheslav Demin, this opens up prospects for the application of local learning rules based on memristors when solving artificial intelligence problems with the use of complex spiking neural network architectures.

Here’s a link to and a citation for the paper,

Yttria-stabilized zirconia cross-point memristive devices for neuromorphic applications by A. V. Emelyanov, K. E. Nikiruy, A. Demin, V. V. Rylkov, A. I. Belov, D. S. Korolev, E. G. Gryaznov, D. A. Pavlov, O. N. Gorshkov, A. N. Mikhaylov, P. Dimitrakis. Microelectronic Engineering Volume 215, 15 July 2019, 110988 First available online 16 May 2019

This paper is behind a paywall.