Tag Archives: brainlike computing

Enabling visible-light programming of ferroelectric memristors for neuromorphic (brainlike) computing

Leave a reply

Hats off to whomever made this illustration. Informative and easy to follow.

Artificial optoelectronic synapses inspired by the human visual system. (a) Schematic diagram of the human visual system. (b) Optically controlled ferroelectric artificial synapse. (c) Optical detector response. (d) Optical synaptic response. (e) Reservoir computing neural network system. (Image: Reproduced with permission from Wiley-VCH Verlag) [downloaded from https://www.nanowerk.com/spotlight/spotid=68593.php]

Michael Berger’s January 30, 2026 Nanowerk Spotlight article describes the problem,

Ferroelectric materials hold innate appeal for engineers designing next-generation computing hardware. Their internal electric dipoles can flip between stable orientations when nudged by an external field, a property called bistability that provides a natural mechanism for storing information without continuous power.

This characteristic makes them promising candidates for memristors, devices whose adjustable electrical resistance can emulate the synapses that connect biological neurons. Unlike conventional computers, which waste time and energy shuttling data between separate memory and processing units, memristor-based systems could store and compute information in the same physical location, much as the brain does.

Lithium niobate has attracted particular attention among ferroelectric candidates. Telecommunications engineers already prize this crystalline compound for its exceptional optical properties, and its responsiveness to both electrical and light signals suggests that artificial synapses might be programmed with photons rather than electrons. Optical control could offer advantages in speed, spatial precision, and energy efficiency that electrical switching cannot match.

But lithium niobate has resisted practical optical programming. Flipping its polarization demands substantial energy, far more than gentle visible light can supply. Laboratory demonstrations have succeeded only under intense illumination or ultraviolet exposure.

Worse, any polarization changes induced by modest visible light decay rapidly once the light source is removed. Under these constraints, lithium niobate functions as a transient optical detector rather than a true nonvolatile memory element.

A paper published in Advanced Materials (“Zn2+ Engineered Low‐Barrier LiNbO3 Enables Visible‐Light Programmable Ferroelectric Memristors for Noise‐Immune Neuromorphic Vision”) by researchers at Hebei University in China reports a materials-engineering strategy that overcomes this barrier. …

Berger’s January 30, 2026 article provides some details about the proposed solution,

… By introducing zinc ions at a concentration of 5 mol.% into the lithium niobate crystal lattice, the team reduced the energy required for polarization switching by approximately 69%. This modification enabled reliable programming using a 405 nm laser at intensities as low as 10 mW cm⁻², a power level comparable to modest indoor lighting. The devices also responded to wavelengths spanning from 405 to 650 nm, and the switched state persists indefinitely after illumination ends.

….

Across 100 switching cycles, the voltage required to change states varied by only 2.2% to 3.2%, uniformity that compares favorably with other reported memristor technologies. The devices retained programmed states for more than 10⁴ seconds and survived over 10⁸ switching cycles without degradation.

These memristors could reliably maintain 24 distinct resistance levels, providing ample gradations for analog computing applications. The ratio between high and low resistance states reached approximately 10³, a thousand-fold difference sufficient for distinguishing stored information with high fidelity.

Under optical stimulation at 405 nm, the devices reproduced several characteristic behaviors of biological synapses. When two light pulses arrived in quick succession, the second pulse triggered a stronger response than the first, mimicking paired-pulse facilitation observed in living neural tissue.

The memristors also captured the distinction between short-term and long-term memory. …

The team demonstrated associative learning resembling Pavlovian conditioning, where initially unrelated stimuli become linked through repeated pairing. The devices also replicated a learning-forgetting-relearning cycle that mirrors human cognitive patterns, showing that relearning previously encountered information required fewer light pulses than initial learning.

…

My excerpts do not do full justice to Berger’s January 30, 2026 article.

For the deeply curious, here’s a link to and a citation for the paper,

Zn2+ Engineered Low-Barrier LiNbO3 Enables Visible-Light Programmable Ferroelectric Memristors for Noise-Immune Neuromorphic Vision by Yifei Pei, Yufei Shang, Gongjie Liu, Shuohua Ma, Mengya Guo, Chao Liu, Yue Hou, Fu Wang, Jianning Wang, Dingxin Liu, Jianhui Zhao, Jianxin Guo, Xiaobing Yan. Advanced Materials Volume 38, Issue 13 3 March 2026 e11352 DOI: https://doi.org/10.1002/adma.202511352 First published online: 22 January 2026

This paper is behind a paywall.

To accelerate self-organizing maps with stacked memristor arrays: compute Euclidean distance in memory

Leave a reply

Michael Berger gives a good description of neuromorphic (brainlike) computing and memristors before (pun alert!) explaining how Euclidean distance factors into the equation in his January 21, 2026 Nanowerk Spotlight article, Note: Links have been removed,

Neuromorphic computing seeks to replicate one of the brain’s most remarkable features: its ability to store information and process it within the same physical structure. Conventional computers, built on the von Neumann architecture that has dominated since the 1940s, continuously shuttle data between separate memory and processing units. This arrangement creates bottlenecks that waste both time and energy, a problem that grows increasingly severe as artificial intelligence demands more computational power.

Memristors, devices whose electrical resistance can be programmed to hold different values, emerged as a promising solution. When arranged in crossbar arrays, memristors can perform vector-matrix multiplication in parallel, with data stored directly in the resistance states of the devices themselves. This capability has enabled impressive demonstrations of supervised neural networks running directly on memristive hardware.

Unsupervised learning algorithms, however, present a different challenge. Unlike supervised learning, where networks train on labeled examples with known correct answers, unsupervised learning discovers hidden patterns and groupings in data without any guidance. This makes it invaluable for exploring datasets where labeling would be prohibitively expensive or simply impossible, such as astronomical surveys containing billions of objects, genomic sequences from unstudied organisms, or sensor streams from industrial equipment.

Self-organizing maps, introduced by Finnish researcher Teuvo Kohonen in the 1980s, represent one of the most elegant approaches to unsupervised learning. These networks transform complex, high-dimensional data into intuitive two-dimensional maps that preserve relationships between similar items.

Imagine trying to understand a dataset where each sample has dozens or hundreds of measurements. A self-organizing map learns to arrange neurons on a grid such that similar inputs activate nearby neurons while dissimilar inputs activate distant ones. The result resembles a topographic map where proximity reflects similarity.

…

And now for Euclidean distance, from the January 21, 2026 Nanowerk Spotlight article, Note: A link has been removed,

Researchers have applied self-organizing maps to cluster gene expression profiles, detect fraud in financial transactions, segment customers by purchasing behavior, and solve optimization puzzles like finding efficient delivery routes.

The algorithm works through competitive learning. When presented with an input, the network identifies which neuron’s stored weights most closely resemble that input. This winning neuron, called the best matching unit, then adjusts along with its neighbors to better represent the input. The network gradually organizes itself through repeated exposure to data.

Finding the best matching unit requires computing Euclidean distance, the straight-line separation between points in multidimensional space. This calculation demands subtraction followed by squaring and accumulation. Memristor crossbars excel at multiplication and addition but cannot natively perform subtraction. Previous hardware implementations have relied on peripheral digital processors or approximate similarity measures, adding complexity and eroding efficiency gains.
A study published in Advanced Functional Materials (“In‐Memory Euclidean Distance Computation in a Stacked Memristor Crossbar for Hardware Self‐Organizing Maps”) by researchers at Hanyang University in Seoul demonstrates a fundamentally different approach. The team developed a three-dimensional stacked memristor crossbar array that computes squared Euclidean distance directly within the memory structure, requiring no external arithmetic circuits.

The innovation exploits a shared middle electrode connecting two vertically stacked memristor layers. The lower layer stores weight values while the upper layer stores their squares. When input voltages are applied, currents flow through both layers simultaneously but with opposite polarity. The middle electrode naturally sums these opposing currents, yielding a net value proportional to squared Euclidean distance. The neuron producing the smallest current corresponds to the best matching unit.

…

For the more technical details, there’s the January 21, 2026 Nanowerk Spotlight article or you can follow this link (with citation),

In-Memory Euclidean Distance Computation in a Stacked Memristor Crossbar for Hardware Self-Organizing Maps by Jinwoo Park, Hyungjin Kim. Advanced Functional Materials Volume 36, Issue 34 27 April 2026 e31235 DOI: https://doi.org/10.1002/adfm.202531235 First published online: 14 January 2026

This paper is open access.

As researchers build high-yield memristor chips, brain-inspired AI moves closer

Leave a reply

This November 7, 2025 news item on Nanowerk highlights research into memristors from Daegu Gyeongbuk Institute of Science and Technology (DGIST), Note: Links have been removed,

A research team at DGIST in South Korea has taken a major step toward building chips that work more like the human brain. Led by Professor Sanghyeon Choi from the Department of Electrical Engineering and Computer Science, the group has developed a highly integrated memristor device at full wafer scale, a milestone for next-generation AI hardware (Nature Communications, “Wafer-scale fabrication of memristive passive crossbar circuits for brain-scale neuromorphic computing”).

…

A November 28, 2025 Daegu Gyeongbuk Institute of Science and Technology (DGIST) press release (also on EurekAlert but published November 4, 2025), which originated the news item, announced its ‘wafer-scale memristor’ and its possible impact on neuromorphic (brain-like) computing,

A research team led by Professor Sanghyeon Choi from the Department of Electrical Engineering and Computer Science at DGIST (President Kunwoo Lee) successfully developed the “memristor,” which is gaining recognition as a next-generation semiconductor device, through mass-integration at the wafer scale. This study proposes a new technological platform for implementing a highly integrated AI semiconductor replicating the human brain, overcoming the limitations of conventional semiconductors.

The human brain contains about 100 billion neurons and around 100 trillion synapses, allowing it to store and process enormous amounts of information within a compact space. Next-generation AI research aims to develop “brain-like AI chips” that replicate this structure. Yet, current AI semiconductors remain far less efficient than the human brain, largely because of their intricate circuitry and substantial power requirements.

The memristor is an emerging alternative option that can overcome these limitations. As a semiconductor device capable of remembering the amount of current flowed, it simultaneously executes memory and computation tasks. Owing to its simple architecture, the circuit can be configured with a much higher density than typical semiconductors. Specifically, an arrangement in a crossbar format enables dozens of times more information to be stored in the same area, compared to SRAM.

However, memristor integration technology has so far been limited to small-scale experimental demonstrations. The main reasons include process complexity, low yield (product completion rate), voltage loss, and current leakage, all of which have hindered its expansion to large-scale wafer production.

Thus, Professor Choi and his team carried out joint research with Dr. Dmitri Strukov’s group at UC Santa Barbara [University of California at Santa Barbara] and introduced a new approach of “co-designing material, component, circuit, and algorithm.” This method enabled the implementation of a memristor crossbar circuit that achieved an approximately 95% yield on a 4-inch wafer without requiring a complex fabrication process.

Furthermore, the research team successfully demonstrated a 3D vertical stacking structure. This signifies the possibility of a memristor-based circuit being expanded into a large-scale AI computation system in the future.

In addition, when a spiking neural network was applied based on the proposed technology, notable efficiency and stable execution were confirmed in actual AI computation.

Professor Choi stated, “This study proposed a method for improving memristor integration technology, which had been limited in the past” and added, “We are expecting it to lead to the development of a next-generation semiconductor platform in the future.”

This study was supported by the U.S. National Science Foundation, Industrial Innovation Talent Growth Support Program of the Korea Institute for Advancement of Technology, and Engineering Academic Research Support Program of the National Research Foundation of Korea’s Science. The research, led by Professor Choi of DGIST as both the first and corresponding author, with Professor Dmitri Strukov of UC Santa Barbara as a co-author, was published in October in the prestigious multidisciplinary journal “Nature Communications.”

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

Wafer-scale fabrication of memristive passive crossbar circuits for brain-scale neuromorphic computing by Sanghyeon Choi, Sai Sukruth Bezugam, Tinish Bhattacharya, Dongseok Kwon & Dmitri B. Strukov. Nature Communications volume 16, Article number: 8757 (2025) Published: 01 October 2025 Version of record: 01 October 2025 DOI: https://doi.org/10.1038/s41467-025-63831-2

This paper is open access.

Ultrafast neuromorphic (brainlike) computing at room temperature made possible by utilizing polariton nonlinearities

Leave a reply

A June 6, 2025 news item on Nanowerk describes research into the development of ultrafast neuromorphic (brainlike) computing, Note: A link has been removed,

Neuromorphic computing, inspired by the human brain, is considered as the next-generation paradigm for artificial intelligence (AI), offering dramatically increased speed and lower energy consumption. While software-based artificial neural networks (ANNs) have made remarkable strides, unlocking their full potential calls for physical platforms that combine ultrafast operation, high computational density, energy efficiency, and scalability.

Among various physical systems, microcavity exciton polaritons have attracted attention for neuromorphic computing due to their ultrafast dynamics, strong nonlinearities, and light-based architecture, which naturally align with the requirements of brain-inspired computation. However, their practical use has been hampered by the need for cryogenic operation and intricate fabrication processes.

In a new paper published in eLight (“Ultrafast neuromorphic computing driven by polariton nonlinearities”), a team of scientists led by Professor Qihua Xiong from Tsinghua University and Beijing Academy of Quantum Information Sciences report a demonstration of neuromorphic computing utilizing perovskite microcavity exciton polaritons operating at room temperature. Their novel system achieves high-speed digit recognition with 92% accuracy using only single-step training and opens new opportunities for scalable, light-driven neural hardware.

…

A June 4, 2025 Light Publishing Center, Changchun Institute of Optics, Fine Mechanics And Physics, CAS (Chinese Academy of Sciences) press release on EurekAlert, which originated the news item, provides more technical details,

The core of their system is a planar FAPbBr3 perovskite microcavity which supports exciton-polariton condensation under non-resonant optical pumping. Input images from the MNIST dataset are optically encoded by a spatial light modulator (SLM) and projected onto the microcavity as spatially structured excitation beams. The resulting polariton emission patterns serve as the output of the ANN, which is then linearly processed using ridge regression. Remarkably, this scheme requires no predefined network structure—only the physical response of the polariton system—and achieves competitive accuracy using a lightweight training set of 900 images.

“Unlike conventional approaches that rely on prefabricated structures or predefined network nodes, our method employs a fully connected spatial mapping, utilizing the entire perovskite sample area without additional structural constraints,” the corresponding author Qihua Xiong replied. This not only improves the system’s scalability but also simplifies experimental realization.

What makes this system stand out is the intrinsic nonlinear and dynamical response of the polaritons. The researchers show that below the condensation threshold, the system behaves nearly linearly, while near and above threshold, nonlinearities emerge sharply, enhancing pattern discrimination. Moreover, by applying ultrafast Kerr-gated time-resolved photoluminescence, the team probes the temporal evolution of polariton responses. They find that polariton dynamics unfold on the picosecond scale and exhibit time-dependent nonlinear mappings, which significantly broaden the system’s capacity for processing complex and temporally varying inputs.

The researchers conclude that “perovskite microcavity exciton polaritons offer ultrafast processing speeds on the picosecond timescale and exhibit exceptionally strong nonlinear interactions, significantly surpassing those in traditional photonic systems.” These attributes make them powerful candidates for future physical neural networks capable of real-time, energy-efficient AI.

This work highlights the growing role of halide perovskites in next-generation photonic computing and marks an important step toward developing all-optical neuromorphic hardware—free from the energy and speed limitations of traditional electronics.

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

Ultrafast neuromorphic computing driven by polariton nonlinearities by Yusong Gan, Ying Shi, Sanjib Ghosh, Haiyun Liu, Huawen Xu & Qihua Xiong. eLight volume 5, Article number: 9 (2025) DOI: https://doi.org/10.1186/s43593-025-00087-9 Published: 02 June 2025

This paper is open access.

A Multidisciplinary Centre for Neuromorphic (brainlike) Computing in the UK

Leave a reply

A May 6, 2025 Aston University press release (also on EurekAlert but published May 7, 2025) announces a UK ‘neuromorphic initiative’, Note: Links have been removed,

  • Aston University to lead the UK’s new centre to pioneer brain-inspired, energy-efficient computing technologies 
  • The initiative will receive £5.6 million over four years from the Engineering and Physical Sciences Research Council [EPSRC]
  • The aim of the centre is to become a focal point for networking and collaboration on fundamental research and technology.

The UK will be getting a new centre to pioneer brain-inspired, energy-efficient computing technologies.

The UK Multidisciplinary Centre for Neuromorphic Computing is led by Aston University and will receive £5.6 million over four years from the UKRI [UK Research and Innovation] Engineering and Physical Sciences Research Council (EPSRC).

The aim of the centre is to become a focal point for networking and collaboration on fundamental research and technology of neuromorphic computing to address the sustainability challenges facing today’s digital infrastructure and artificial intelligence systems.

The centre will be led by the Aston Institute of Photonic Technologies (AIPT) and will include the world-leading researchers from Aston University, the University of Oxford, the University of Cambridge, the University of Southampton, Queen Mary University of London, Loughborough University and the University of Strathclyde. 

Neuromorphic computing seeks to replicate the brain’s structural and functional principles, however scientists currently lack a deep, system-level understanding of how the human brain computes at cellular and network scales. The researchers aim to tackle that challenge directly, blending stem-cell-derived human neuron experiments with advanced computational models, low-power algorithms and novel photonic hardware.

The centre team includes world-leading researchers with broad and complementary expertise in neuroscience, non-conventional computing algorithms, photonics, opto- and nano-electronics and materials science. In collaboration with policymakers and industrial partners the scientists and engineers aim to demonstrate the capabilities of neuromorphic computing across a range of sectors and applications. The centre will be supported by a broad network of industry partners including Microsoft Research, Thales, BT, QinetiQ, Nokia Bell Labs, Hewlett Packard Labs, Leonardo, Northrop Grumman and a number of small to medium enterprises. Their contribution will focus on enhancing the centre’s impact on society.

Professor Rhein Parri, co-director and neurophysiologist at Aston University said: “For the first time, we can combine the study of living human neurons with that of advanced computing platforms to co-develop the future of computing. 

“This project is an exciting leap forward, learning from biology and technology in ways that were not previously possible.”

The experts aim to co-design brain-inspired neuromorphic systems by studying human neuronal function using the latest human induced pluripotent stem cell – or hiPSC technologies – and developing new computational paradigms and low-power AI algorithms. They also plan to create devices and hardware that are inspired by biological systems, like the human brain. These devices will use light – or photonic hardware – to process information. This approach will be the next big step in making computing more energy-efficient and capable of handling many tasks at the same time. They also aim to create a sustainable UK research ecosystem through training, road mapping, and international collaboration.

Professor Sergei K. Turitsyn, director of the centre and AIPT, said: “The project’s ambition is not only to develop future technologies, but also to create a new internationally known UK research brand in neuromorphic computing that will unite the UK’s best minds across disciplines and will lead to sustainable operation and a long-term impact. It’s a proud moment for AIPT and Aston University to lead this national effort.”

Professor Natalia Berloff, co-director of the centre who is based at the University of Cambridge said: “One of the most exciting aspects of neuromorphic computing is the potential of photonic hardware to deliver truly brain-like efficiency. 

“Light-based processors can exploit massive parallelism and ultrafast signal propagation to outperform conventional electronics on demanding AI workloads, while consuming far less power. By combining these photonic architectures with insights from living human neurons, we aim to co-design neuromorphic systems that move beyond incremental improvements and toward a genuinely transformative computing paradigm.”

In addition, the researchers aim to tackle the increasing global energy footprint of information and communication technologies which is developing at an unsustainable pace, driven partly by the explosive growth of artificial intelligence. Today’s AI systems are built on traditional computing hardware with increasingly high-power consumption (kW), posing a barrier to scalability and sustainability. In contrast, the human brain performs complex computation and communication tasks using just 20 watts.

Professor Dimitra Georgiadou, co-director of the centre who is based at the University of Southampton added: “To address the challenge of substantially lowering the power consumption in electronics, novel materials and device architectures are needed that can effectively emulate computation in the brain and cellular responses to certain stimuli.”

The centre’s ambition goes beyond technology development as it aims to serve as a foundation for a long-term, interdisciplinary research ecosystem – actively expanding its membership and reach over time. It aims to establish a sustainable centre that continues to be a focal point for the community and will thrive beyond the initial funding period, reinforcing innovation, partnership, and impact in the field of neuromorphic computing.

…

Good luck to this effort to lower power consumption.

‘Super-Turing AI’ uses less energy to mimic brain

Leave a reply

Neuromorphic (brainlike) engineering and neuromorphic computing being long time interests here, this March 26, 2025 new item on ScienceDaily caught my eye,

Artificial Intelligence (AI) can perform complex calculations and analyze data faster than any human, but to do so requires enormous amounts of energy. The human brain is also an incredibly powerful computer, yet it consumes very little energy.

As technology companies increasingly expand, a new approach to AI’s “thinking,” developed by researchers including Texas A&M University engineers, mimics the human brain and has the potential to revolutionize the AI industry.

…

A March 25, 2025 Texas A&M University news release (also on EurekAlert) by Lesley Henton, which originated the news item, delves further into the creation of a “Super-Turing AI,” Note: Links have been removed,

As technology companies increasingly expand, a new approach to AI’s “thinking,” developed by researchers including Texas A&M University engineers, mimics the human brain and has the potential to revolutionize the AI industry.

Dr. Suin Yi, assistant professor of electrical and computer engineering at Texas A&M’s College of Engineering, is on a team of researchers that developed “Super-Turing AI,” which operates more like the human brain. This new AI integrates certain processes instead of separating them and then migrating huge amounts of data like current systems do.

The Energy Crisis In AI

Today’s AI systems, including large language models [LLM] such as OpenAI [a company not an LLM] and ChatGPT [an LLM produced by OpenAI], require immense computing power and are housed in expansive data centers that consume vast amounts of electricity.

“These data centers are consuming power in gigawatts, whereas our brain consumes 20 watts,” Suin explained. “That’s 1 billion watts compared to just 20. Data centers that are consuming this energy are not sustainable with current computing methods. So while AI’s abilities are remarkable, the hardware and power generation needed to sustain it is still needed.”

The substantial energy demands not only escalate operational costs but also raise environmental concerns, given the carbon footprint associated with large-scale data centers. As AI becomes more integrated, addressing its sustainability becomes increasingly critical.

Emulating The Brain

Yi and team believe the key to solving this problem lies in nature — specifically, the human brain’s neural processes.

In the brain, the functions of learning and memory are not separated, they are integrated. Learning and memory rely on connections between neurons, called “synapses,” where signals are transmitted. Learning strengthens or weakens synaptic connections through a process called “synaptic plasticity,” forming new circuits and altering existing ones to store and retrieve information. 

By contrast, in current computing systems, training (how the AI is taught) and memory (data storage) happen in two separate places within the computer hardware. Super-Turing AI is revolutionary because it bridges this efficiency gap, so the computer doesn’t have to migrate enormous amounts of data from one part of its hardware to another.

“Traditional AI models rely heavily on backpropagation — a method used to adjust neural networks during training,” Yi said. “While effective, backpropagation is not biologically plausible and is computationally intensive.

“What we did in that paper is troubleshoot the biological implausibility present in prevailing machine learning algorithms,” he said. “Our team explores mechanisms like Hebbian learning and spike-timing-dependent plasticity — processes that help neurons strengthen connections in a way that mimics how real brains learn.”

Hebbian learning principles are often summarized as “cells that fire together, wire together.” This approach aligns more closely with how neurons in the brain strengthen their connections based on activity patterns. By integrating such biologically inspired mechanisms, the team aims to develop AI systems that require less computational power without compromising performance.

In a test, a circuit using these components helped a drone navigate a complex environment — without prior training — learning and adapting on the fly. This approach was faster, more efficient and used less energy than traditional AI.

Why This Matters For The Future Of AI

This research could be a game-changer for the AI industry. Companies are racing to build larger and more powerful AI models, but their ability to scale is limited by hardware and energy constraints. In some cases, new AI applications require building entire new data centers, further increasing environmental and economic costs.

Yi emphasizes that innovation in hardware is just as crucial as advancements in AI systems themselves. “Many people say AI is just a software thing, but without computing hardware, AI cannot exist,” he said.

Looking Ahead: Sustainable AI Development

Super-Turing AI represents a pivotal step toward sustainable AI development. By reimagining AI architectures to mirror the efficiency of the human brain, the industry can address both economic and environmental challenges.

Yi and his team hope that their research will lead to a new generation of AI that is both smarter and more efficient.

“Modern AI like ChatGPT is awesome, but it’s too expensive. We’re going to make sustainable AI,” Yi said. “Super-Turing AI could reshape how AI is built and used, ensuring that as it continues to advance, it does so in a way that benefits both people and the planet.”

There’s no mention of a memristor but there is a ‘synaptic resistor’, which I find puzzling. Is a synaptic resistor something different? In a search with these search terms “synaptic resistor memristor” I found this,

The term “memristive synapses” signifies the amalgamation of memristor functionality with synaptic characteristics, resulting in a novel approach to neuromorphic computing.

I’m guessing memristive synapses can also be called synaptic resistors or, at the least, are related concepts.

I pulled the definition from,

Resistive Switching Properties in Memristors for Optoelectronic Synaptic Memristors: Deposition Techniques, Key Performance Parameters, and Applications by Rajwali Khan, Naveed Ur Rehman, Shahid Iqbal, Sherzod Abdullaev, and Haila M. Aldosari. ACS Applied Electronic Materials Vol 6/ Issue 1 pp. 73–119 DOI: https://doi.org/10.1021/acsaelm.3c01323 Published December 29, 2023 Copyright © 2023 The Authors. Published by American Chemical Society. This publication is licensed under
CC-BY 4.0

Getting back to this latest work from Texas A&M University, here’s a link to and a citation for Dr. Suin Yi and his team’s paper,

HfZrO-based synaptic resistor circuit for a Super-Turing intelligent system by Jungmin Lee, Rahul Shenoy, Atharva Deo, Suin Yi, Dawei Gao, David Qiao, Mingjie Xu, Shiva Asapu, Zixuan Rong, Dhruva Nathan, Yong Hei, Dharma Paladugu, Jian-Guo Zheng, J. Joshua Yang, R. Stanley Williams, Qing Wu, and Yong Chen. Science Advances 28 Feb 2025 Vol 11, Issue 9 DOI: 10.1126/sciadv.adr2082

This paper is open access.

Notice that one of the Super Turing paper’s authors is R. Stanley Williams who ‘discovered’ the memristor in 2008. You can read his November 28, 2008 article “How We Found the Missing Memristor; The memristor—the functional equivalent of a synapse—could revolutionize circuit design” in the IEEE Spectrum online,

It’s time to stop shrinking. Moore’s Law, the semiconductor industry’s obsession with the shrinking of transistors and their commensurate steady doubling on a chip about every two years, has been the source of a 50-year technical and economic revolution. Whether this scaling paradigm lasts for five more years or 15, it will eventually come to an end. The emphasis in electronics design will have to shift to devices that are not just increasingly infinitesimal but increasingly capable.

Earlier this year, I and my colleagues at Hewlett-Packard Labs, in Palo Alto, Calif., surprised the electronics community with a fascinating candidate for such a device: the memristor. It had been theorized nearly 40 years ago, but because no one had managed to build one, it had long since become an esoteric curiosity. That all changed on 1 May [2008], when my group published the details of the memristor in Nature.

…

For anyone interested in a trip down memory road, I have a few comments from the theorist (Leon Chua) mentioned in his 2008 article in this April 13, 2010 posting (scroll down to the ‘More on memristors’ subhead).

Next-generation neuromorphic, semiconductor-based, ultra-small computing chip learns and corrects itself

Leave a reply

This is yet another of my memristor posts. Researchers from Korea Advanced Institute of Science and Technology (KAIST) have some exciting news according to a January 21, 2025 news item on ScienceDaily,

Existing computer systems have separate data processing and storage devices, making them inefficient for processing complex data like AI. A KAIST research team has developed a memristor-based integrated system similar to the way our brain processes information. It is now ready for application in various devices including smart security cameras, allowing them to recognize suspicious activity immediately without having to rely on remote cloud servers, and medical devices with which it can help analyze health data in real time.

KAIST (President Kwang Hyung Lee) announced on the 17th of January [2025] that the joint research team of Professor Shinhyun Choi and Professor Young-Gyu Yoon of the School of Electrical Engineering has developed a next-generation neuromorphic semiconductor-based ultra-small computing chip that can learn and correct errors on its own.

…

A January 17, 2025 KAIST press release (also on EurekAlert but published January 20, 2025), which originated the news item, provides more information,

What is special about this computing chip is that it can learn and correct errors that occur due to non-ideal characteristics that were difficult to solve in existing neuromorphic devices. For example, when processing a video stream, the chip learns to automatically separate a moving object from the background, and it becomes better at this task over time.

This self-learning ability has been proven by achieving accuracy comparable to ideal computer simulations in real-time image processing. The research team’s main achievement is that it has completed a system that is both reliable and practical, beyond the development of brain-like components.

The research team has developed the world’s first memristor-based integrated system that can adapt to immediate environmental changes, and has presented an innovative solution that overcomes the limitations of existing technology.

…

At the heart of this innovation is a next-generation semiconductor device called a memristor*. The variable resistance characteristics of this device can replace the role of synapses in neural networks, and by utilizing it, data storage and computation can be performed simultaneously, just like our brain cells.

*Memristor: A compound word of memory and resistor, next-generation electrical device whose resistance value is determined by the amount and direction of charge that has flowed between the two terminals in the past.

The research team designed a highly reliable memristor that can precisely control resistance changes and developed an efficient system that excludes complex compensation processes through self-learning. This study is significant in that it experimentally verified the commercialization possibility of a next-generation neuromorphic semiconductor-based integrated system that supports real-time learning and inference.

This technology will revolutionize the way artificial intelligence is used in everyday devices, allowing AI tasks to be processed locally without relying on remote cloud servers, making them faster, more privacy-protected, and more energy-efficient.

“This system is like a smart workspace where everything is within arm’s reach instead of having to go back and forth between desks and file cabinets,” explained KAIST researchers Hakcheon Jeong and Seungjae Han, who led the development of this technology. “This is similar to the way our brain processes information, where everything is processed efficiently at once at one spot.”

The research was conducted with Hakcheon Jeong and Seungjae Han, the students of Integrated Master’s and Doctoral Program at KAIST School of Electrical Engineering being the co-first authors, the results of which was published online in the international academic journal, Nature Electronics, on January 8, 2025.

…

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

Self-supervised video processing with self-calibration on an analogue computing platform based on a selector-less memristor array by Hakcheon Jeong, Seungjae Han, See-On Park, Tae Ryong Kim, Jongmin Bae, Taehwan Jang, Yoonho Cho, Seokho Seo, Hyun-Jun Jeong, Seungwoo Park, Taehoon Park, Juyoung Oh, Jeongwoo Park, Kwangwon Koh, Kang-Ho Kim, Dongsuk Jeon, Inyong Kwon, Young-Gyu Yoon & Shinhyun Choi. Nature Electronics volume 8, pages 168–178 (2025) DOI: https://doi.org/10.1038/s41928-024-01318-6 Published: 08 January 2025 Issue Date: February 2025

This paper is behind a paywall.

Neuromorphic wires (inspired by nerve cells) amplify their own signals—no amplifiers needed

Leave a reply

Katherine Bourzac’s September 16, 2024 article for the IEEE (Institute for Electrical and Electronics Engineers) Spectrum magazine provides an accessible (relatively speaking) description of a possible breakthrough for neuromorphic computing, Note: Links have been removed,

In electrical engineering, “we just take it for granted that the signal decays” as it travels, says Timothy Brown, a postdoc in materials physics at Sandia National Lab who was part of the group of researchers who made the self-amplifying device. Even the best wires and chip interconnects put up resistance to the flow of electrons, degrading signal quality over even relatively small distances. This constrains chip designs—lossy interconnects are broken up into ever smaller lengths, and signals are bolstered by buffers and drivers. A 1-square-centimeter chip has about 10,000 repeaters to drive signals, estimates R. Stanley Williams, a professor of computer engineering at Texas A&M University.

Williams is one of the pioneers of neuromorphic computing, which takes inspiration from the nervous system. Axons, the electrical cables that carry signals from the body of a nerve cell to synapses where they connect with projections from other cells, are made up of electrically resistant materials. Yet they can carry high fidelity signals over long distances. The longest axons in the human body are about 1 meter, running from the base of the spine to the feet. Blue whales are thought to have 30 m long axons stretching to the tips of their tails. If something bites the whale’s tail, it will react rapidly. Even from 30 meters away, “the pulses arrive perfectly,” says Williams. “That’s something that doesn’t exist in electrical engineering.”

That’s because axons are active transmission lines: they provide gain to the signal along their length. Williams says he started pondering how to mimic this in an inorganic system 12 years ago. A grant from the US Department of Energy enabled him to build a team with the necessary resources to make it happen. The team included Williams, Brown, and Suhas Kumar, a materials physicist at Sandia.

Axons are coated with an insulating layer called the myelin sheath. Where there are gaps in the sheath, negatively charged sodium ions and positively charged potassium ions can move in and out of the axon, changing the voltage across the cell membrane and pumping in energy in the process. Some of that energy gets taken up by the electrical signal, amplifying it.

…

Williams and his team wanted to mimic this in a simple structure. They didn’t try to mimic all the physical structures in axons—instead, they sought guidance in a mathematical description of how they amplify signals. Axons operate in a mode called the “edge of chaos,” which combines stable and unstable qualities. This may seem inherently contradictory. Brown likens this kind of system to a saddle that’s curved with two dips. The saddle curves up towards the front and the back, keeping you stable as you rock back and forth. But if you get jostled from side to side, you’re more likely to fall off. When you’re riding in the saddle, you’re operating at the edge of chaos, in a semistable state. In the abstract space of electrical engineering, that jostling is equivalent to wiggles in current and voltage.

…

There’s a long way to go from this first experimental demonstration to a reimagining of computer chip interconnects. The team is providing samples for other researchers [emphasis mine] who want to verify their measurements. And they’re trying other materials to see how well they do—LaCoO3 [lanthanum colbalt oxide] is only the first one they’ve tested.

Williams hopes this research will show electrical engineers new ideas about how to move forward. “The dream is to redesign chips,” he says. Electrical engineers have long known about nonlinear dynamics, but have hardly ever taken advantage of them, Williams says. “This requires thinking about things and doing measurements differently than they have been done for 50 years,” he says.

If you have the time, please read Bourzac’s September 16, 2024 article in its entirety. For those who want the technical nitty gritty, here’s a link to and a citation for the paper,

Axon-like active signal transmission by Timothy D. Brown, Alan Zhang, Frederick U. Nitta, Elliot D. Grant, Jenny L. Chong, Jacklyn Zhu, Sritharini Radhakrishnan, Mahnaz Islam, Elliot J. Fuller, A. Alec Talin, Patrick J. Shamberger, Eric Pop, R. Stanley Williams & Suhas Kumar. Nature volume 633, pages 804–810 (2024) DOI: https://doi.org/10.1038/s41586-024-07921 Published online: 11 September 2024 Issue Date: 26 September 2024

This paper is open access.

Light-based neural networks

Leave a reply

It’s unusual to see the same headline used to highlight research from two different teams released in such proximity, February 2024 and July 2024, respectively. Both of these are neuromorphic (brainlike) computing stories.

February 2024: Neural networks made of light

The first team’s work is announced in a February 21, 2024 Friedrich Schiller University press release, Note: A link has been removed,

Researchers from the Leibniz Institute of Photonic Technology (Leibniz IPHT) and the Friedrich Schiller University in Jena, along with an international team, have developed a new technology that could significantly reduce the high energy demands of future AI systems. This innovation utilizes light for neuronal computing, inspired by the neural networks of the human brain. It promises not only more efficient data processing but also speeds many times faster than current methods, all while consuming considerably less energy. Published in the prestigious journal „Advanced Science,“ their work introduces new avenues for environmentally friendly AI applications, as well as advancements in computerless diagnostics and intelligent microscopy.

Artificial intelligence (AI) is pivotal in advancing biotechnology and medical procedures, ranging from cancer diagnostics to the creation of new antibiotics. However, the ecological footprint of large-scale AI systems is substantial. For instance, training extensive language models like ChatGPT-3 requires several gigawatt-hours of energy—enough to power an average nuclear power plant at full capacity for several hours.

Prof. Mario Chemnitz, new Junior Professor of Intelligent Photonic SystemsExternal link at Friedrich Schiller University Jena, and Dr Bennet Fischer from Leibniz IPHT in Jena, in collaboration with their international team, have devised an innovative method to develop potentially energy-efficient computing systems that forego the need for extensive electronic infrastructure. They harness the unique interactions of light waves within optical fibers to forge an advanced artificial learning system.

A single fiber instead of thousands of components

Unlike traditional systems that rely on computer chips containing thousands of electronic components, their system uses a single optical fiber. This fiber is capable of performing the tasks of various neural networks—at the speed of light. “We utilize a single optical fiber to mimic the computational power of numerous neural networks,“ Mario Chemnitz, who is also leader of the “Smart Photonics“ junior research group at Leibniz IPHT, explains. “By leveraging the unique physical properties of light, this system will enable the rapid and efficient processing of vast amounts of data in the future.“

Delving into the mechanics reveals how information transmission occurs through the mixing of light frequencies: Data—whether pixel values from images or frequency components of an audio track—are encoded onto the color channels of ultrashort light pulses. These pulses carry the information through the fiber, undergoing various combinations, amplifications, or attenuations. The emergence of new color combinations at the fiber’s output enables the prediction of data types or contexts. For example, specific color channels can indicate visible objects in images or signs of illness in a voice.

A prime example of machine learning is identifying different numbers from thousands of handwritten characters. Mario Chemnitz, Bennet Fischer, and their colleagues from the Institut National de la Recherche Scientifique (INRS) in Québec utilized their technique to encode images of handwritten digits onto light signals and classify them via the optical fiber. The alteration in color composition at the fiber’s end forms a unique color spectrum—a „fingerprint“ for each digit. Following training, the system can analyze and recognize new handwriting digits with significantly reduced energy consumption.

System recognizes COVID-19 from voice samples

“In simpler terms, pixel values are converted into varying intensities of primary colors—more red or less blue, for instance,“ Mario Chemnitz details. “Within the fiber, these primary colors blend to create the full spectrum of the rainbow. The shade of our mixed purple, for example, reveals much about the data processed by our system.“

The team has also successfully applied this method in a pilot study to diagnose COVID-19 infections using voice samples, achieving a detection rate that surpasses the best digital systems to date.

“We are the first to demonstrate that such a vibrant interplay of light waves in optical fibers can directly classify complex information without any additional intelligent software,“ Mario Chemnitz states.

Since December 2023, Mario Chemnitz has held the position of Junior Professor of Intelligent Photonic Systems at Friedrich Schiller University Jena. Following his return from INRS in Canada in 2022, where he served as a postdoc, Chemnitz has been leading an international team at Leibniz IPHT in Jena. With Nexus funding support from the Carl Zeiss Foundation, their research focuses on exploring the potentials of non-linear optics. Their goal is to develop computer-free intelligent sensor systems and microscopes, as well as techniques for green computing.

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

Neuromorphic Computing via Fission-based Broadband Frequency Generation by Bennet Fischer, Mario Chemnitz, Yi Zhu, Nicolas Perron, Piotr Roztocki, Benjamin MacLellan, Luigi Di Lauro, A. Aadhi, Cristina Rimoldi, Tiago H. Falk, Roberto Morandotti. Advanced Science Volume 10, Issue 35 December 15, 2023 2303835 DOI: https://doi.org/10.1002/advs.202303835. First published: 02 October 2023

This paper is open access.

July 2024: Neural networks made of light

A July 12, 2024 news item on ScienceDaily announces research from another German team,

Scientists propose a new way of implementing a neural network with an optical system which could make machine learning more sustainable in the future. The researchers at the Max Planck Institute for the Science of Light have published their new method in Nature Physics, demonstrating a method much simpler than previous approaches.

…

A July 12, 2024 Max Planck Institute for the Science of Light press release (also on EurekAlert), which originated the news item, provides more detail about their approach to neuromorphic computiing,

Machine learning and artificial intelligence are becoming increasingly widespread with applications ranging from computer vision to text generation, as demonstrated by ChatGPT. However, these complex tasks require increasingly complex neural networks; some with many billion parameters. This rapid growth of neural network size has put the technologies on an unsustainable path due to their exponentially growing energy consumption and training times. For instance, it is estimated that training GPT-3 consumed more than 1,000 MWh of energy, which amounts to the daily electrical energy consumption of a small town. This trend has created a need for faster, more energy- and cost-efficient alternatives, sparking the rapidly developing field of neuromorphic computing. The aim of this field is to replace the neural networks on our digital computers with physical neural networks. These are engineered to perform the required mathematical operations physically in a potentially faster and more energy-efficient way.

Optics and photonics are particularly promising platforms for neuromorphic computing since energy consumption can be kept to a minimum. Computations can be performed in parallel at very high speeds only limited by the speed of light. However, so far, there have been two significant challenges: Firstly, realizing the necessary complex mathematical computations requires high laser powers. Secondly, the lack of an efficient general training method for such physical neural networks.

Both challenges can be overcome with the new method proposed by Clara Wanjura and Florian Marquardt from the Max Planck Institute for the Science of Light in their new article in Nature Physics. “Normally, the data input is imprinted on the light field. However, in our new methods we propose to imprint the input by changing the light transmission,” explains Florian Marquardt, Director at the Institute. In this way, the input signal can be processed in an arbitrary fashion. This is true even though the light field itself behaves in the simplest way possible in which waves interfere without otherwise influencing each other. Therefore, their approach allows one to avoid complicated physical interactions to realize the required mathematical functions which would otherwise require high-power light fields. Evaluating and training this physical neural network would then become very straightforward: “It would really be as simple as sending light through the system and observing the transmitted light. This lets us evaluate the output of the network. At the same time, this allows one to measure all relevant information for the training”, says Clara Wanjura, the first author of the study. The authors demonstrated in simulations that their approach can be used to perform image classification tasks with the same accuracy as digital neural networks.

In the future, the authors are planning to collaborate with experimental groups to explore the implementation of their method. Since their proposal significantly relaxes the experimental requirements, it can be applied to many physically very different systems. This opens up new possibilities for neuromorphic devices allowing physical training over a broad range of platforms.

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

Fully nonlinear neuromorphic computing with linear wave scattering by Clara C. Wanjura & Florian Marquardt. Nature Physics (2024) DOI: https://doi.org/10.1038/s41567-024-02534-9 Published: 09 July 2024

This paper is open access.

Proposed platform for brain-inspired computing

Leave a reply

Researchers at the University of California at Santa Barbara (UCSB) have proposed a more energy-efficient architecture for neuromorphic (brainlike or brain-inspored) computing according to a June 25, 2024 news item on ScienceDaily,

Computers have come so far in terms of their power and potential, rivaling and even eclipsing human brains in their ability to store and crunch data, make predictions and communicate. But there is one domain where human brains continue to dominate: energy efficiency.

“The most efficient computers are still approximately four orders of magnitude — that’s 10,000 times — higher in energy requirements compared to the human brain for specific tasks such as image processing and recognition, although they outperform the brain in tasks like mathematical calculations,” said UC Santa Barbara electrical and computer engineering Professor Kaustav Banerjee, a world expert in the realm of nanoelectronics. “Making computers more energy efficient is crucial because the worldwide energy consumption by on-chip electronics stands at #4 in the global rankings of nation-wise energy consumption, and it is increasing exponentially each year, fueled by applications such as artificial intelligence.” Additionally, he said, the problem of energy inefficient computing is particularly pressing in the context of global warming, “highlighting the urgent need to develop more energy-efficient computing technologies.”

….

A June 24, 2024 UCSB news release (also on Eurekalert), which originated the news item, delves further into the subject,

Neuromorphic (NM) computing has emerged as a promising way to bridge the energy efficiency gap. By mimicking the structure and operations of the human brain, where processing occurs in parallel across an array of low power-consuming neurons, it may be possible to approach brain-like energy efficiency. In a paper published in the journal Nature Communications, Banerjee and co-workers Arnab Pal, Zichun Chai, Junkai Jiang and Wei Cao, in collaboration with researchers Vivek De and Mike Davies from Intel Labs propose such an ultra-energy efficient platform, using 2D transition metal dichalcogenide (TMD)-based tunnel-field-effect transistors (TFETs). Their platform, the researchers say, can bring the energy requirements to within two orders of magnitude (about 100 times) with respect to the human brain.

Leakage currents and subthreshold swing

The concept of neuromorphic computing has been around for decades, though the research around it has intensified only relatively recently. Advances in circuitry that enable smaller, denser arrays of transistors, and therefore more processing and functionality for less power consumption are just scratching the surface of what can be done to enable brain-inspired computing. Add to that an appetite generated by its many potential applications, such as AI and the Internet-of-Things, and it’s clear that expanding the options for a hardware platform for neuromorphic computing must be addressed in order to move forward.

Enter the team’s 2D tunnel-transistors. Emerging out of Banerjee’s longstanding research efforts to develop high-performance, low-power consumption transistors to meet the growing hunger for processing without a matching increase in power requirement, these atomically thin, nanoscale transistors are responsive at low voltages, and as the foundation of the researchers’ NM platform, can mimic the highly energy efficient operations of the human brain. In addition to lower off-state currents, the 2D TFETs also have a low subthreshold swing (SS), a parameter that describes how effectively a transistor can switch from off to on. According to Banerjee, a lower SS means a lower operating voltage, and faster and more efficient switching.

“Neuromorphic computing architectures are designed to operate with very sparse firing circuits,” said lead author Arnab Pal, “meaning they mimic how neurons in the brain fire only when necessary.” In contrast to the more conventional von Neumann architecture of today’s computers, in which data is processed sequentially, memory and processing components are separated and which continuously draw power throughout the entire operation, an event-driven system such as a NM computer fires up only when there is input to process, and memory and processing are distributed across an array of transistors. Companies like Intel and IBM have developed brain-inspired platforms, deploying billions of interconnected transistors and generating significant energy savings.

However, there’s still room for energy efficiency improvement, according to the researchers.

“In these systems, most of the energy is lost through leakage currents when the transistors are off, rather than during their active state,” Banerjee explained. A ubiquitous phenomenon in the world of electronics, leakage currents are small amounts of electricity that flow through a circuit even when it is in the off state (but still connected to power). According to the paper, current NM chips use traditional metal-oxide-semiconductor field-effect transistors (MOSFETs) which have a high on-state current, but also high off-state leakage. “Since the power efficiency of these chips is constrained by the off-state leakage, our approach — using tunneling transistors with much lower off-state current — can greatly improve power efficiency,” Banerjee said.

When integrated into a neuromorphic circuit, which emulates the firing and reset of neurons, the TFETs proved themselves more energy efficient than state-of-the-art MOSFETs, particularly the FinFETs (a MOSFET design that incorporates vertical “fins” as a way to provide better control of switching and leakage). TFETs are still in the experimental stage, however the performance and energy efficiency of neuromorphic circuits based on them makes them a promising candidate for the next generation of brain-inspired computing.

According to co-authors Vivek De (Intel Fellow) and Mike Davies (Director of Intel’s Neuromorphic Computing Lab), “Once realized, this platform can bring the energy consumption in chips to within two orders of magnitude with respect to the human brain — not accounting for the interface circuitry and memory storage elements. This represents a significant improvement from what is achievable today.”

Eventually, one can realize three-dimensional versions of these 2D-TFET based neuromorphic circuits to provide even closer emulation of the human brain, added Banerjee, widely recognized as one of the key visionaries behind 3D integrated circuits that are now witnessing wide scale commercial proliferation.

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

An ultra energy-efficient hardware platform for neuromorphic computing enabled by 2D-TMD tunnel-FETs by Arnab Pal, Zichun Chai, Junkai Jiang, Wei Cao, Mike Davies, Vivek De & Kaustav Banerjee. Nature Communications volume 15, Article number: 3392 (2024) DOI: https://doi.org/10.1038/s41467-024-46397-3 Published: 22 April 2024

This paper is open access.