Tag Archives: neuronal activity

Brainy and brainy: a novel synaptic architecture and a neuromorphic computing platform called SpiNNaker

I have two items about brainlike computing. The first item hearkens back to memristors, a topic I have been following since 2008. (If you’re curious about the various twists and turns just enter  the term ‘memristor’ in this blog’s search engine.) The latest on memristors is from a team than includes IBM (US), École Politechnique Fédérale de Lausanne (EPFL; Swizterland), and the New Jersey Institute of Technology (NJIT; US). The second bit comes from a Jülich Research Centre team in Germany and concerns an approach to brain-like computing that does not include memristors.

Multi-memristive synapses

In the inexorable march to make computers function more like human brains (neuromorphic engineering/computing), an international team has announced its latest results in a July 10, 2018 news item on Nanowerk,

Two New Jersey Institute of Technology (NJIT) researchers, working with collaborators from the IBM Research Zurich Laboratory and the École Polytechnique Fédérale de Lausanne, have demonstrated a novel synaptic architecture that could lead to a new class of information processing systems inspired by the brain.

The findings are an important step toward building more energy-efficient computing systems that also are capable of learning and adaptation in the real world. …

A July 10, 2018 NJIT news release (also on EurekAlert) by Tracey Regan, which originated by the news item, adds more details,

The researchers, Bipin Rajendran, an associate professor of electrical and computer engineering, and S. R. Nandakumar, a graduate student in electrical engineering, have been developing brain-inspired computing systems that could be used for a wide range of big data applications.

Over the past few years, deep learning algorithms have proven to be highly successful in solving complex cognitive tasks such as controlling self-driving cars and language understanding. At the heart of these algorithms are artificial neural networks – mathematical models of the neurons and synapses of the brain – that are fed huge amounts of data so that the synaptic strengths are autonomously adjusted to learn the intrinsic features and hidden correlations in these data streams.

However, the implementation of these brain-inspired algorithms on conventional computers is highly inefficient, consuming huge amounts of power and time. This has prompted engineers to search for new materials and devices to build special-purpose computers that can incorporate the algorithms. Nanoscale memristive devices, electrical components whose conductivity depends approximately on prior signaling activity, can be used to represent the synaptic strength between the neurons in artificial neural networks.

While memristive devices could potentially lead to faster and more power-efficient computing systems, they are also plagued by several reliability issues that are common to nanoscale devices. Their efficiency stems from their ability to be programmed in an analog manner to store multiple bits of information; however, their electrical conductivities vary in a non-deterministic and non-linear fashion.

In the experiment, the team showed how multiple nanoscale memristive devices exhibiting these characteristics could nonetheless be configured to efficiently implement artificial intelligence algorithms such as deep learning. Prototype chips from IBM containing more than one million nanoscale phase-change memristive devices were used to implement a neural network for the detection of hidden patterns and correlations in time-varying signals.

“In this work, we proposed and experimentally demonstrated a scheme to obtain high learning efficiencies with nanoscale memristive devices for implementing learning algorithms,” Nandakumar says. “The central idea in our demonstration was to use several memristive devices in parallel to represent the strength of a synapse of a neural network, but only chose one of them to be updated at each step based on the neuronal activity.”

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

Neuromorphic computing with multi-memristive synapses by Irem Boybat, Manuel Le Gallo, S. R. Nandakumar, Timoleon Moraitis, Thomas Parnell, Tomas Tuma, Bipin Rajendran, Yusuf Leblebici, Abu Sebastian, & Evangelos Eleftheriou. Nature Communications volume 9, Article number: 2514 (2018) DOI: https://doi.org/10.1038/s41467-018-04933-y Published 28 June 2018

This is an open access paper.

Also they’ve got a couple of very nice introductory paragraphs which I’m including here, (from the June 28, 2018 paper in Nature Communications; Note: Links have been removed),

The human brain with less than 20 W of power consumption offers a processing capability that exceeds the petaflops mark, and thus outperforms state-of-the-art supercomputers by several orders of magnitude in terms of energy efficiency and volume. Building ultra-low-power cognitive computing systems inspired by the operating principles of the brain is a promising avenue towards achieving such efficiency. Recently, deep learning has revolutionized the field of machine learning by providing human-like performance in areas, such as computer vision, speech recognition, and complex strategic games1. However, current hardware implementations of deep neural networks are still far from competing with biological neural systems in terms of real-time information-processing capabilities with comparable energy consumption.

One of the reasons for this inefficiency is that most neural networks are implemented on computing systems based on the conventional von Neumann architecture with separate memory and processing units. There are a few attempts to build custom neuromorphic hardware that is optimized to implement neural algorithms2,3,4,5. However, as these custom systems are typically based on conventional silicon complementary metal oxide semiconductor (CMOS) circuitry, the area efficiency of such hardware implementations will remain relatively low, especially if in situ learning and non-volatile synaptic behavior have to be incorporated. Recently, a new class of nanoscale devices has shown promise for realizing the synaptic dynamics in a compact and power-efficient manner. These memristive devices store information in their resistance/conductance states and exhibit conductivity modulation based on the programming history6,7,8,9. The central idea in building cognitive hardware based on memristive devices is to store the synaptic weights as their conductance states and to perform the associated computational tasks in place.

The two essential synaptic attributes that need to be emulated by memristive devices are the synaptic efficacy and plasticity. …

It gets more complicated from there.

Now onto the next bit.


At a guess, those capitalized N’s are meant to indicate ‘neural networks’. As best I can determine, SpiNNaker is not based on the memristor. Moving on, a July 11, 2018 news item on phys.org announces work from a team examining how neuromorphic hardware and neuromorphic software work together,

A computer built to mimic the brain’s neural networks produces similar results to that of the best brain-simulation supercomputer software currently used for neural-signaling research, finds a new study published in the open-access journal Frontiers in Neuroscience. Tested for accuracy, speed and energy efficiency, this custom-built computer named SpiNNaker, has the potential to overcome the speed and power consumption problems of conventional supercomputers. The aim is to advance our knowledge of neural processing in the brain, to include learning and disorders such as epilepsy and Alzheimer’s disease.

A July 11, 2018 Frontiers Publishing news release on EurekAlert, which originated the news item, expands on the latest work,

“SpiNNaker can support detailed biological models of the cortex–the outer layer of the brain that receives and processes information from the senses–delivering results very similar to those from an equivalent supercomputer software simulation,” says Dr. Sacha van Albada, lead author of this study and leader of the Theoretical Neuroanatomy group at the Jülich Research Centre, Germany. “The ability to run large-scale detailed neural networks quickly and at low power consumption will advance robotics research and facilitate studies on learning and brain disorders.”

The human brain is extremely complex, comprising 100 billion interconnected brain cells. We understand how individual neurons and their components behave and communicate with each other and on the larger scale, which areas of the brain are used for sensory perception, action and cognition. However, we know less about the translation of neural activity into behavior, such as turning thought into muscle movement.

Supercomputer software has helped by simulating the exchange of signals between neurons, but even the best software run on the fastest supercomputers to date can only simulate 1% of the human brain.

“It is presently unclear which computer architecture is best suited to study whole-brain networks efficiently. The European Human Brain Project and Jülich Research Centre have performed extensive research to identify the best strategy for this highly complex problem. Today’s supercomputers require several minutes to simulate one second of real time, so studies on processes like learning, which take hours and days in real time are currently out of reach.” explains Professor Markus Diesmann, co-author, head of the Computational and Systems Neuroscience department at the Jülich Research Centre.

He continues, “There is a huge gap between the energy consumption of the brain and today’s supercomputers. Neuromorphic (brain-inspired) computing allows us to investigate how close we can get to the energy efficiency of the brain using electronics.”

Developed over the past 15 years and based on the structure and function of the human brain, SpiNNaker — part of the Neuromorphic Computing Platform of the Human Brain Project — is a custom-built computer composed of half a million of simple computing elements controlled by its own software. The researchers compared the accuracy, speed and energy efficiency of SpiNNaker with that of NEST–a specialist supercomputer software currently in use for brain neuron-signaling research.

“The simulations run on NEST and SpiNNaker showed very similar results,” reports Steve Furber, co-author and Professor of Computer Engineering at the University of Manchester, UK. “This is the first time such a detailed simulation of the cortex has been run on SpiNNaker, or on any neuromorphic platform. SpiNNaker comprises 600 circuit boards incorporating over 500,000 small processors in total. The simulation described in this study used just six boards–1% of the total capability of the machine. The findings from our research will improve the software to reduce this to a single board.”

Van Albada shares her future aspirations for SpiNNaker, “We hope for increasingly large real-time simulations with these neuromorphic computing systems. In the Human Brain Project, we already work with neuroroboticists who hope to use them for robotic control.”

Before getting to the link and citation for the paper, here’s a description of SpiNNaker’s hardware from the ‘Spiking neural netowrk’ Wikipedia entry, Note: Links have been removed,

Neurogrid, built at Stanford University, is a board that can simulate spiking neural networks directly in hardware. SpiNNaker (Spiking Neural Network Architecture) [emphasis mine], designed at the University of Manchester, uses ARM processors as the building blocks of a massively parallel computing platform based on a six-layer thalamocortical model.[5]

Now for the link and citation,

Performance Comparison of the Digital Neuromorphic Hardware SpiNNaker and the Neural Network Simulation Software NEST for a Full-Scale Cortical Microcircuit Model by
Sacha J. van Albada, Andrew G. Rowley, Johanna Senk, Michael Hopkins, Maximilian Schmidt, Alan B. Stokes, David R. Lester, Markus Diesmann, and Steve B. Furber. Neurosci. 12:291. doi: 10.3389/fnins.2018.00291 Published: 23 May 2018

As noted earlier, this is an open access paper.

Transparent graphene electrode technology and complex brain imaging

Michael Berger has written a May 24, 2018 Nanowerk Spotlight article about some of the latest research on transparent graphene electrode technology and the brain (Note: A link has been removed),

In new work, scientists from the labs of Kuzum [Duygu Kuzum, an Assistant Professor of Electrical and Computer Engineering at the University of California, San Diego {UCSD}] and Anna Devor report a transparent graphene microelectrode neural implant that eliminates light-induced artifacts to enable crosstalk-free integration of 2-photon microscopy, optogenetic stimulation, and cortical recordings in the same in vivo experiment. The new class of transparent brain implant is based on monolayer graphene. It offers a practical pathway to investigate neuronal activity over multiple spatial scales extending from single neurons to large neuronal populations.

Conventional metal-based microelectrodes cannot be used for simultaneous measurements of multiple optical and electrical parameters, which are essential for comprehensive investigation of brain function across spatio-temporal scales. Since they are opaque, they block the field of view of the microscopes and generate optical shadows impeding imaging.

More importantly, they cause light induced artifacts in electrical recordings, which can significantly interfere with neural signals. Transparent graphene electrode technology presented in this paper addresses these problems and allow seamless and crosstalk-free integration of optical and electrical sensing and manipulation technologies.

In their work, the scientists demonstrate that by careful design of key steps in the fabrication process for transparent graphene electrodes, the light-induced artifact problem can be mitigated and virtually artifact-free local field potential (LFP) recordings can be achieved within operating light intensities.

“Optical transparency of graphene enables seamless integration of imaging, optogenetic stimulation and electrical recording of brain activity in the same experiment with animal models,” Kuzum explains. “Different from conventional implants based on metal electrodes, graphene-based electrodes do not generate any electrical artifacts upon interacting with light used for imaging or optogenetics. That enables crosstalk free integration of three modalities: imaging, stimulation and recording to investigate brain activity over multiple spatial scales extending from single neurons to large populations of neurons in the same experiment.”

The team’s new fabrication process avoids any crack formation in the transfer process, resulting in a 95-100% yield for the electrode arrays. This fabrication quality is important for expanding this technology to high-density large area transparent arrays to monitor brain-scale cortical activity in large animal models or humans.

“Our technology is also well-suited for neurovascular and neurometabolic studies, providing a ‘gold standard’ neuronal correlate for optical measurements of vascular, hemodynamic, and metabolic activity,” Kuzum points out. “It will find application in multiple areas, advancing our understanding of how microscopic neural activity at the cellular scale translates into macroscopic activity of large neuron populations.”

“Combining optical techniques with electrical recordings using graphene electrodes will allow to connect the large body of neuroscience knowledge obtained from animal models to human studies mainly relying on electrophysiological recordings of brain-scale activity,” she adds.

Next steps for the team involve employing this technology to investigate coupling and information transfer between different brain regions.

This work is part of the US BRAIN (Brain Research through Advancing Innovative Neurotechnologies) initiative and there’s more than one team working with transparent graphene electrodes. John Hewitt in an Oct. 21, 2014 posting on ExtremeTech describes two other teams’ work (Note: Links have been removed),

The solution [to the problems with metal electrodes], now emerging from multiple labs throughout the universe is to build flexible, transparent electrode arrays from graphene. Two studies in the latest issue of Nature Communications, one from the University of Wisconsin-Madison and the other from Penn [University of Pennsylvania], describe how to build these devices.

The University of Wisconsin researchers are either a little bit smarter or just a little bit richer, because they published their work open access. It’s a no-brainer then that we will focus on their methods first, and also in more detail. To make the arrays, these guys first deposited the parylene (polymer) substrate on a silicon wafer, metalized it with gold, and then patterned it with an electron beam to create small contact pads. The magic was to then apply four stacked single-atom-thick graphene layers using a wet transfer technique. These layers were then protected with a silicon dioxide layer, another parylene layer, and finally molded into brain signal recording goodness with reactive ion etching.

PennTransparentelectrodeThe researchers went with four graphene layers because that provided optimal mechanical integrity and conductivity while maintaining sufficient transparency. They tested the device in opto-enhanced mice whose neurons expressed proteins that react to blue light. When they hit the neurons with a laser fired in through the implant, the protein channels opened and fired the cell beneath. The masterstroke that remained was then to successfully record the electrical signals from this firing, sit back, and wait for the Nobel prize office to call.

The Penn State group [Note: Every reearcher mentioned in the paper Hewitt linked to is from the University of Pennsylvania] in the  used a similar 16-spot electrode array (pictured above right), and proceeded — we presume — in much the same fashion. Their angle was to perform high-resolution optical imaging, in particular calcium imaging, right out through the transparent electrode arrays which simultaneously recorded in high-temporal-resolution signals. They did this in slices of the hippocampus where they could bring to bear the complex and multifarious hardware needed to perform confocal and two-photon microscopy. These latter techniques provide a boost in spatial resolution by zeroing in over narrow planes inside the specimen, and limiting the background by the requirement of two photons to generate an optical signal. We should mention that there are voltage sensitive dyes available, in addition to standard calcium dyes, which can almost record the fastest single spikes, but electrical recording still reigns supreme for speed.

What a mouse looks like with an optogenetics system plugged in

What a mouse looks like with an optogenetics system plugged in

One concern of both groups in making these kinds of simultaneous electro-optic measurements was the generation of light-induced artifacts in the electrical recordings. This potential complication, called the Becqueral photovoltaic effect, has been known to exist since it was first demonstrated back in 1839. When light hits a conventional metal electrode, a photoelectrochemical (or more simply, a photovoltaic) effect occurs. If present in these recordings, the different signals could be highly disambiguatable. The Penn researchers reported that they saw no significant artifact, while the Wisconsin researchers saw some small effects with their device. In particular, when compared with platinum electrodes put into the opposite side cortical hemisphere, the Wisconsin researchers found that the artifact from graphene was similar to that obtained from platinum electrodes.

Here’s a link to and a citation for the latest research from UCSD,

Deep 2-photon imaging and artifact-free optogenetics through transparent graphene microelectrode arrays by Martin Thunemann, Yichen Lu, Xin Liu, Kıvılcım Kılıç, Michèle Desjardins, Matthieu Vandenberghe, Sanaz Sadegh, Payam A. Saisan, Qun Cheng, Kimberly L. Weldy, Hongming Lyu, Srdjan Djurovic, Ole A. Andreassen, Anders M. Dale, Anna Devor, & Duygu Kuzum. Nature Communicationsvolume 9, Article number: 2035 (2018) doi:10.1038/s41467-018-04457-5 Published: 23 May 2018

This paper is open access.

You can find out more about the US BRAIN initiative here and if you’re curious, you can find out more about the project at UCSD here. Duygu Kuzum (now at UCSD) was at  the University of Pennsylvania in 2014 and participated in the work mentioned in Hewitt’s 2014 posting.

Bidirectional prosthetic-brain communication with light?

The possibility of not only being able to make a prosthetic that allows a tetraplegic to grab a coffee but to feel that coffee  cup with their ‘hand’ is one step closer to reality according to a Feb. 22, 2017 news item on ScienceDaily,

Since the early seventies, scientists have been developing brain-machine interfaces; the main application being the use of neural prosthesis in paralyzed patients or amputees. A prosthetic limb directly controlled by brain activity can partially recover the lost motor function. This is achieved by decoding neuronal activity recorded with electrodes and translating it into robotic movements. Such systems however have limited precision due to the absence of sensory feedback from the artificial limb. Neuroscientists at the University of Geneva (UNIGE), Switzerland, asked whether it was possible to transmit this missing sensation back to the brain by stimulating neural activity in the cortex. They discovered that not only was it possible to create an artificial sensation of neuroprosthetic movements, but that the underlying learning process occurs very rapidly. These findings, published in the scientific journal Neuron, were obtained by resorting to modern imaging and optical stimulation tools, offering an innovative alternative to the classical electrode approach.

A Feb. 22, 2017 Université de Genève press release on EurekAlert, which originated the news item, provides more detail,

Motor function is at the heart of all behavior and allows us to interact with the world. Therefore, replacing a lost limb with a robotic prosthesis is the subject of much research, yet successful outcomes are rare. Why is that? Until this moment, brain-machine interfaces are operated by relying largely on visual perception: the robotic arm is controlled by looking at it. The direct flow of information between the brain and the machine remains thus unidirectional. However, movement perception is not only based on vision but mostly on proprioception, the sensation of where the limb is located in space. “We have therefore asked whether it was possible to establish a bidirectional communication in a brain-machine interface: to simultaneously read out neural activity, translate it into prosthetic movement and reinject sensory feedback of this movement back in the brain”, explains Daniel Huber, professor in the Department of Basic Neurosciences of the Faculty of Medicine at UNIGE.

Providing artificial sensations of prosthetic movements

In contrast to invasive approaches using electrodes, Daniel Huber’s team specializes in optical techniques for imaging and stimulating brain activity. Using a method called two-photon microscopy, they routinely measure the activity of hundreds of neurons with single cell resolution. “We wanted to test whether mice could learn to control a neural prosthesis by relying uniquely on an artificial sensory feedback signal”, explains Mario Prsa, researcher at UNIGE and the first author of the study. “We imaged neural activity in the motor cortex. When the mouse activated a specific neuron, the one chosen for neuroprosthetic control, we simultaneously applied stimulation proportional to this activity to the sensory cortex using blue light”. Indeed, neurons of the sensory cortex were rendered photosensitive to this light, allowing them to be activated by a series of optical flashes and thus integrate the artificial sensory feedback signal. The mouse was rewarded upon every above-threshold activation, and 20 minutes later, once the association learned, the rodent was able to more frequently generate the correct neuronal activity.

This means that the artificial sensation was not only perceived, but that it was successfully integrated as a feedback of the prosthetic movement. In this manner, the brain-machine interface functions bidirectionally. The Geneva researchers think that the reason why this fabricated sensation is so rapidly assimilated is because it most likely taps into very basic brain functions. Feeling the position of our limbs occurs automatically, without much thought and probably reflects fundamental neural circuit mechanisms. This type of bidirectional interface might allow in the future more precisely displacing robotic arms, feeling touched objects or perceiving the necessary force to grasp them.

At present, the neuroscientists at UNIGE are examining how to produce a more efficient sensory feedback. They are currently capable of doing it for a single movement, but is it also possible to provide multiple feedback channels in parallel? This research sets the groundwork for developing a new generation of more precise, bidirectional neural prostheses.

Towards better understanding the neural mechanisms of neuroprosthetic control

By resorting to modern imaging tools, hundreds of neurons in the surrounding area could also be observed as the mouse learned the neuroprosthetic task. “We know that millions of neural connections exist. However, we discovered that the animal activated only the one neuron chosen for controlling the prosthetic action, and did not recruit any of the neighbouring neurons”, adds Daniel Huber. “This is a very interesting finding since it reveals that the brain can home in on and specifically control the activity of just one single neuron”. Researchers can potentially exploit this knowledge to not only develop more stable and precise decoding techniques, but also gain a better understanding of most basic neural circuit functions. It remains to be discovered what mechanisms are involved in routing signals to the uniquely activated neuron.

Caption: A novel optical brain-machine interface allows bidirectional communication with the brain. While a robotic arm is controlled by neuronal activity recorded with optical imaging (red laser), the position of the arm is fed back to the brain via optical microstimulation (blue laser). Credit: © Daniel Huber, UNIGE

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

Rapid Integration of Artificial Sensory Feedback during Operant Conditioning of Motor Cortex Neurons by Mario Prsa, Gregorio L. Galiñanes, Daniel Huber. Neuron Volume 93, Issue 4, p929–939.e6, 22 February 2017 DOI: http://dx.doi.org/10.1016/j.neuron.2017.01.023 Open access funded by European Research Council

This paper is open access.