Tag Archives: Los Alamos National Laboratory (LANL)

D-Wave Systems demonstrates quantum advantage on optimization problems with a 5,000-qubit programmable spin glass

Leave a reply

This May 17, 2023 article by Ingrid Fadelli for phys.org describes quantum research performed by D-Wave Systems (a company in Vancouver, Canada) and Boston University (Massachusetts, US), Note: Links have been removed,

Over the past decades, researchers and companies worldwide have been trying to develop increasingly advanced quantum computers. The key objective of their efforts is to create systems that will outperform classical computers on specific tasks, which is also known as realizing “quantum advantage.”

A research team at D-Wave Inc., a quantum computing company, recently created a new quantum computing system that outperforms classical computing systems on optimization problems. This system, introduced in a paper in Nature, is based on a programmable spin glass with 5,000 qubits (the quantum equivalents of bits in classical computing).

“This work validates the original hypothesis behind quantum annealing, coming full circle from some seminal experiments conducted in the 1990s,” Andrew D. King, one of the researchers who carried out the study, told Phys.org.

“These original experiments took chunks of spin-glass alloy and subjected them to varying magnetic fields, and the observations suggested that if we made a programmable quantum spin glass, it could drive down to low-energy states of optimization problems faster than analogous classical algorithms. A Science paper published in 2014 tried to verify this on a D-Wave Two processor, but no speedup was found.”

…

“This is a ‘full circle’ moment, in the sense that we have verified and extended the hypothesis of the UChicago [University of Chicago] and NEC [Nippon Electric Company] researchers; quantum annealing shows a scaling advantage over simulated thermal annealing,” King said. “Ours is the largest programmable quantum simulation ever performed; reproducing it classically is way beyond the reach of known methods.”

…

“We have a clear view of quantum effects and very clear evidence, both theoretical and experimental, that the quantum effects are conferring a computational scaling advantage over classical methods,” King said. “We want to highlight the difference between this original definition of quantum advantage and the fact that it is sometimes used as a stand-in term for quantum supremacy, which we have not demonstrated. [emphases mine] Gate-model quantum computers have not shown any capabilities approaching this for optimization, and I personally don’t believe they ever will.”

…

“For a long time, it was subject for debate whether or not coherent quantum dynamics were playing any role at all in quantum annealing,” King said. “While this controversy has been rebuked by previous works, this new research is the clearest demonstration yet, by far.”

…

An April 19, 2023 D-Wave Systems news release, which seems to have been the basis for Fadelli’s article, provides more detail in a release that functions as a research announcement and a sales tool, Note: Links have been removed,

D-Wave Quantum Inc. (NYSE: QBTS), a leader in quantum computing systems, software, and services—and the only provider building both annealing and gate-model quantum computers, today published a peer-reviewed milestone paper showing the performance of its 5,000 qubit Advantage™ quantum computer is significantly faster than classical compute on 3D spin glass optimization problems, an intractable class of optimization problems. This paper also represents the largest programmable quantum simulation reported to date.

The paper—a collaboration between scientists from D-Wave and Boston University—entitled “Quantum critical dynamics in a 5,000-qubit programmable spin glass,” was published in the peer-reviewed journal Nature today and is available here. Building upon research conducted on up to 2,000 qubits last September, the study shows that the D-Wave quantum processor can compute coherent quantum dynamics in large-scale optimization problems. This work was done using D-Wave’s commercial-grade annealing-based quantum computer, which is accessible for customers to use today.

With immediate implications to optimization, the findings show that coherent quantum annealing can improve solution quality faster than classical algorithms. The observed speedup matches the theory of coherent quantum annealing and shows​ a direct connection between coherence and the core computational power of quantum annealing.

“This research marks a significant achievement for quantum technology, as it demonstrates a computational advantage over classical approaches for an intractable class of optimization problems,” said Dr. Alan Baratz, CEO of D-Wave. “For those seeking evidence of quantum annealing’s unrivaled performance, this work offers definitive proof.

This work supports D-Wave’s ongoing commitment to relentless scientific innovation and product delivery, as the company continues development on its future annealing and gate model quantum computers. To date, D-Wave has brought to market five generations of quantum computers and launched an experimental prototype of its sixth-generation machine, the Advantage2™ system, in June 2022. The full Advantage2 system is expected to feature 7,000+ qubits, 20-way connectivity and higher coherence to solve even larger and more complex problems. Read more about the research in our Medium post here.

Paper’s Authors and Leading Industry Voices Echo Support

“This is an important advance in the study of quantum phase transitions on quantum annealers. It heralds a revolution in experimental many-body physics and bodes well for practical applications of quantum computing,” said Wojciech Zurek, theoretical physicist at Los Alamos National Laboratory and leading authority on quantum theory. Dr. Zurek is widely renowned for his groundbreaking contribution to our understanding of the early universe as well as condensed matter systems through the discovery of the celebrated Kibble-Zurek mechanism. This mechanism underpins the physics behind the experiment reported in this paper. “The same hardware that has already provided useful experimental proving ground for quantum critical dynamics can be also employed to seek low-energy states that assist in finding solutions to optimization problems.”

“Disordered magnets, such as spin glasses, have long functioned as model systems for testing solvers of complex optimization problems,” said Gabriel Aeppli, professor of physics at ETH Zürich and EPF Lausanne, and head of the Photon Science Division of the Paul Scherrer Institut. Professor Aeppli coauthored the first experimental paper demonstrating advantage of quantum annealing over thermal annealing in reaching ground state of disordered magnets. “This paper gives evidence that the quantum dynamics of a dedicated hardware platform are faster than for known classical algorithms to find the preferred, lowest energy state of a spin glass, and so promises to continue to fuel the further development of quantum annealers for dealing with practical problems.”

“As a physicist who has built my career on computer simulations of quantum systems, it has been amazing to experience first-hand the transformative capabilities of quantum annealing devices,” said Anders Sandvik, professor of physics at Boston University and a coauthor of the paper. “This paper already demonstrates complex quantum dynamics on a scale beyond any classical simulation method, and I’m very excited about the expected enhanced performance of future devices. I believe we are now entering an era when quantum annealing becomes an essential tool for research on complex systems.”

“This work marks a major step towards large-scale quantum simulations of complex materials,” said Hidetoshi Nishimori, Professor, Institute of Innovative Research, Tokyo Institute of Technology and one of the original inventors of quantum annealing. “We can now expect novel physical phenomena to be revealed by quantum simulations using quantum annealing, ultimately leading to the design of materials of significant societal value.”

“This represents some of the most important experimental work ever performed in quantum optimization,” said Dr. Andrew King, director of performance research at D-Wave. “We’ve demonstrated a speedup over simulated annealing, in strong agreement with theory, providing high-quality solutions for large-scale problems. This work shows clear evidence of quantum dynamics in optimization, which we believe paves the way for even more complex problem-solving using quantum annealing in the future. The work exhibits a programmable realization of lab experiments that originally motivated quantum annealing 25 years ago.”

“Not only is this the largest demonstration of quantum simulation to date, but it also provides the first experimental evidence, backed by theory, that coherent quantum dynamics can accelerate the attainment of better solutions in quantum annealing,” said Mohammad Amin, fellow, quantum algorithms and systems, at D-Wave. “The observed speedup can be attributed to complex critical dynamics during quantum phase transition, which cannot be replicated by classical annealing algorithms, and the agreement between theory and experiment is remarkable. We believe these findings have significant implications for quantum optimization, with practical applications in addressing real-world problems.”

About D-Wave Quantum Inc.

D-Wave is a leader in the development and delivery of quantum computing systems, software, and services, and is the world’s first commercial supplier of quantum computers—and the only company building both annealing quantum computers and gate-model quantum computers. Our mission is to unlock the power of quantum computing today to benefit business and society. We do this by delivering customer value with practical quantum applications for problems as diverse as logistics, artificial intelligence, materials sciences, drug discovery, scheduling, cybersecurity, fault detection, and financial modeling. D-Wave’s technology is being used by some of the world’s most advanced organizations, including Volkswagen, Mastercard, Deloitte, Davidson Technologies, ArcelorMittal, Siemens Healthineers, Unisys, NEC Corporation, Pattison Food Group Ltd., DENSO, Lockheed Martin, Forschungszentrum Jülich, University of Southern California, and Los Alamos National Laboratory.

Forward-Looking Statements

This press release contains forward-looking statements within the meaning of the Private Securities Litigation Reform Act of 1995, which statements are based on beliefs and assumptions and on information currently available. In some cases, you can identify forward-looking statements by the following words: “may,” “will,” “could,” “would,” “should,” “expect,” “intend,” “plan,” “anticipate,” “believe,” “estimate,” “predict,” “project,” “potential,” “continue,” “ongoing,” or the negative of these terms or other comparable terminology, although not all forward-looking statements contain these words. These statements involve risks, uncertainties, and other factors that may cause actual results, levels of activity, performance, or achievements to be materially different from the information expressed or implied by these forward-looking statements. We caution you that these statements are based on a combination of facts and factors currently known by us and our projections of the future, which are subject to a number of risks. Forward-looking statements in this press release include, but are not limited to, statements regarding the impact of the results of this study; the company’s Advantage2™ experimental prototype; and the potential for future problem-solving using quantum annealing. We cannot assure you that the forward-looking statements in this press release will prove to be accurate. These forward-looking statements are subject to a number of risks and uncertainties, including, among others, various factors beyond management’s control, including general economic conditions and other risks, our ability to expand our customer base and the customer adoption of our solutions, and the uncertainties and factors set forth in the sections entitled “Risk Factors” and “Cautionary Note Regarding Forward-Looking Statements” in D-Wave Quantum Inc.’s Form S-4 Registration Statement, as amended, previously filed with the Securities and Exchange Commission, as well as factors associated with companies, such as D-Wave, that are engaged in the business of quantum computing, including anticipated trends, growth rates, and challenges in those businesses and in the markets in which they operate; the outcome of any legal proceedings that may be instituted against us; risks related to the performance of our business and the timing of expected business or financial milestones; unanticipated technological or project development challenges, including with respect to the cost and or timing thereof; the performance of the our products; the effects of competition on our business; the risk that we will need to raise additional capital to execute our business plan, which may not be available on acceptable terms or at all; the risk that we may never achieve or sustain profitability; the risk that we are unable to secure or protect our intellectual property; volatility in the price of our securities; and the risk that our securities will not maintain the listing on the NYSE. Furthermore, if the forward-looking statements contained in this press release prove to be inaccurate, the inaccuracy may be material. In addition, you are cautioned that past performance may not be indicative of future results. In light of the significant uncertainties in these forward-looking statements, you should not place undue reliance on these statements in making an investment decision or regard these statements as a representation or warranty by any person we will achieve our objectives and plans in any specified time frame, or at all. The forward-looking statements in this press release represent our views as of the date of this press release. We anticipate that subsequent events and developments will cause our views to change. However, while we may elect to update these forward-looking statements at some point in the future, we have no current intention of doing so except to the extent required by applicable law. You should, therefore, not rely on these forward-looking statements as representing our views as of any date subsequent to the date of this press release.

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

Quantum critical dynamics in a 5,000-qubit programmable spin glass by Andrew D. King, Jack Raymond, Trevor Lanting, Richard Harris, Alex Zucca, Fabio Altomare, Andrew J. Berkley, Kelly Boothby, Sara Ejtemaee, Colin Enderud, Emile Hoskinson, Shuiyuan Huang, Eric Ladizinsky, Allison J. R. MacDonald, Gaelen Marsden, Reza Molavi, Travis Oh, Gabriel Poulin-Lamarre, Mauricio Reis, Chris Rich, Yuki Sato, Nicholas Tsai, Mark Volkmann, Jed D. Whittaker, Jason Yao, Anders W. Sandvik & Mohammad H. Amin. Nature volume 617, pages 61–66 (2023) DOI: https://doi.org/10.1038/s41586-023-05867-2 Published: 19 April 2023 Issue Date: 04 May 2023

This paper is behind a paywall but there is an open access version on the arxiv website which means that it has had some peer review but may differ from the version in Nature.

Trying to understand how neural networks work

Leave a reply

It seems no one (not even the experts) really understands how ‘artificial intelligence that learns’ actually works. This is the second time I’ve stumbled onto a similar statement made by experts. Last time (see my July 28, 2022 posting [part 1 of 2] and scroll down to the “A deep dive into AI?” subhead for a quote from a recent Council of Canadian Academies report on AI for Science and Engineering) they were unable to gave a stable definition for artificial intelligence.

Researchers at Los Alamos are looking at new ways to compare neural networks. This image was created with an artificial intelligence software called Stable Diffusion, using the prompt “Peeking into the black box of neural networks.” Courtesy: Los Alamos National Laboratorory

A September 13, 2022 Los Alamos National Laboratory (LANL) news release (also on EurekAlert) provides detail about how researchers are addressing the problem,

A team at Los Alamos National Laboratory has developed a novel approach for comparing neural networks that looks within the “black box” of artificial intelligence to help researchers understand neural network behavior. Neural networks recognize patterns in datasets; they are used everywhere in society, in applications such as virtual assistants, facial recognition systems and self-driving cars.

“The artificial intelligence research community doesn’t necessarily have a complete understanding of what neural networks are doing; they give us good results, but we don’t know how or why,” said Haydn Jones, a researcher in the Advanced Research in Cyber Systems group at Los Alamos. “Our new method does a better job of comparing neural networks, which is a crucial step toward better understanding the mathematics behind AI.”

Jones is the lead author of the paper “If You’ve Trained One You’ve Trained Them All: Inter-Architecture Similarity Increases With Robustness,” which was presented recently at the Conference on Uncertainty in Artificial Intelligence. In addition to studying network similarity, the paper is a crucial step toward characterizing the behavior of robust neural networks.

Neural networks are high performance, but fragile. For example, self-driving cars use neural networks to detect signs. When conditions are ideal, they do this quite well. However, the smallest aberration — such as a sticker on a stop sign — can cause the neural network to misidentify the sign and never stop.

To improve neural networks, researchers are looking at ways to improve network robustness. One state-of-the-art approach involves “attacking” networks during their training process. Researchers intentionally introduce aberrations and train the AI to ignore them. This process is called adversarial training and essentially makes it harder to fool the networks.

Jones, Los Alamos collaborators Jacob Springer and Garrett Kenyon, and Jones’ mentor Juston Moore, applied their new metric of network similarity to adversarially trained neural networks, and found, surprisingly, that adversarial training causes neural networks in the computer vision domain to converge to very similar data representations, regardless of network architecture, as the magnitude of the attack increases.

“We found that when we train neural networks to be robust against adversarial attacks, they begin to do the same things,” Jones said.

There has been extensive effort in industry and in the academic community searching for the “right architecture” for neural networks, but the Los Alamos team’s findings indicate that the introduction of adversarial training narrows this search space substantially. As a result, the AI research community may not need to spend as much time exploring new architectures, knowing that adversarial training causes diverse architectures to converge to similar solutions.

“By finding that robust neural networks are similar to each other, we’re making it easier to understand how robust AI might really work. We might even be uncovering hints as to how perception occurs in humans and other animals,” Jones said.

Should you be curious about future events, the Association for Uncertainty in Artificial Intelligence (AUAI), a non-profit, organizes an annual conference.

The reddest red and Schrödinger’s red pixel

Leave a reply
Caption: Schrödinger’s red pixel by quasi-bound-states in-the-continuum Credit: 123RF

Science keeps moving. First, there was the June 2022 news and, then, there was the August 2022 news.

A June 8, 2022 Agency for Science, Technology and Research (A*STAR) press release (also on EurekAlert but published June 7, 2022 as an ‘article highlight’) announces more research into structural colour along with some colour theory from Erwin Schrödinger,

The brilliant and often iridescent colours that we see in some species of birds, beetles and butterflies arise from a regular arrangement of nanostructures that scatter selective wavelengths of light more strongly to generate colour. These colours are called structural colours, which usually range from blues to greens, and even magenta. However, vibrant or saturated reds are elusive and notably absent from the structural colour range in both natural and synthetic realms.

To achieve highly saturated reds, the material needs to absorb light from all wavelengths shorter than ~600 nm and reflect the remaining longer wavelengths, doing both as completely as possible. This sharp transition from absorption to reflection was prescribed theoretically by none other than Erwin Schrödinger of quantum theory fame. However, the physics of resonators tell us that high-order optical resonances in blue will also occur as soon as we have a fundamental resonance in red. This combination of blue and red thus results in the magenta observed in nature. It is therefore challenging to achieve the Schrödinger’s red pixel, which would produce the most saturated red in the world. Current nanoantenna-based approaches are insufficient to simultaneously satisfy the above conditions.

Researchers from the Agency for Science, Technology and Research’s (A*STAR) Institute of Materials Research and Engineering (IMRE), National University of Singapore (NUS) and Singapore University of Technology and Design (SUTD) have collaborated to design and realise reds at the ultimate limit of saturation as predicted by theory, where the team worked together on conceptualisation methodology, fabrications, characterisations and simulations. This research was published in Science Advances on 23 February 2022.

The design consists of regularly arranged silicon nanoantennas in the shape of ellipses. These produce possibly the most saturated and brightest reds with ~80% reflectance, exceeding the reds in the standard red, green and blue gamut (sRGB) and other well-known red pigments, e.g. cadmium red … .

The nanoantennas support two partially overlapping quasi bound-states-in-the-continuum modes, where the optimal dimensions of the silicon nanoantenna arrays are derived by using a gradient descent algorithm to enable the antennas to achieve sharp spectral edges at red wavelengths. At the same time, high-order modes at blue or green wavelengths are suppressed via engineering the substrate‑induced diffraction channels and the absorption of amorphous silicon.

Potential uses for Schrödinger’s red include developing a polarisation dependent encryption method, with plans to scale up the Schrödinger’s red pixel for applications like functional nanofabrication devices such as optical spectrometers and reflective displays with high colour saturation.

“With this new design that can achieve the most saturated and brightest reds, we can exploit its sensitivity to polarisation and illumination angle on potential applications for information encryption. This proposed concept and design methodology could also be generalised to other Schrödinger’s colour pixels. The highly-saturated red achieved could be potentially scaled up through methods such as deep ultraviolet and nano-imprint lithography, to reach the dimensions of reflective displays based on multilayer film configuration, which could lead to potential applications like compact red filters, highly saturated reflective displays, nonlocal metasurfaces, and miniaturised spectrometers”, said Dr. Dong Zhaogang, Deputy Department Head of Nanofabrication at A*STAR’s IMRE.

“The creation of the record-high saturation and brightness in red opens up possibilities for a plethora of applications related to anti-counterfeiting technologies, high-calibre colour display and more, which were previously perceived as unachievable with structural colour. It showcases a wonderful synergy between conceptual breakthrough, powerful algorithm and advanced nanofabrication”, said Prof. Cheng-Wei Qiu, Dean’s Chair Professor at NUS.

“This work in structural colours goes to show that we can sometimes outdo evolution through clever use of the tools in nanofabrication and accurate optical simulations”, said Prof. Joel Yang, Provost Chair Professor and Associate Professor in Engineering Product Development at SUTD.

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

Schrödinger’s red pixel by quasi-bound-states-in-the-continuum by Zhaogang Dong, Lei Jin, Soroosh Daqiqeh Rezaei, Hao Wang, Yang Chen, Febiana Tjiptoharsono, Jinfa Ho, Sergey Gorelik, Ray Jia Hong Ng, Qifeng Ruan, Cheng-Wei Qiu and Joel K. W. Yang. Science Advances Vol 8, Issue 8 DOI: 10.1126/sciadv.abm4512 Published 23 Feb 2022

This paper is open access.

Math error, colour theory, and perception

An August 10, 2022 news item on phys.org announced a math error made by Erwin Schrödinger and others,

A new study corrects an important error in the 3D mathematical space developed by the Nobel Prize-winning physicist Erwin Schrödinger and others, and used by scientists and industry for more than 100 years to describe how your eye distinguishes one color from another. The research has the potential to boost scientific data visualizations, improve TVs and recalibrate the textile and paint industries.

“The assumed shape of color space requires a paradigm shift,” said Roxana Bujack, a computer scientist with a background in mathematics who creates scientific visualizations at Los Alamos National Laboratory. Bujack is lead author of the paper by a Los Alamos team in the Proceedings of the National Academy of Sciences on the mathematics of color perception.

“Our research shows that the current mathematical model of how the eye perceives color differences is incorrect. That model was suggested by Bernhard Riemann and developed by Hermann von Helmholtz and Erwin Schrödinger—all giants in mathematics and physics—and proving one of them wrong is pretty much the dream of a scientist,” said Bujack.

…

While the Los Alamos National Laboratory work was published in April 2022 (online) and May 2022 (in print), their news announcement doesn’t seem to have been made until August. I can’t be certain but I believe this should have an impact on the work from A*STAR as that team’s paper cites: E. Schrödinger, Theorie der Pigmente von größter Leuchtkraft. Ann. Phys. 367, 603–622 (1920).

An August 10, 2022 Los Alamos National Laboratory (LANL) news release (also on EurekAlert) provides more information about the discovery,

Modeling human color perception enables automation of image processing, computer graphics and visualization tasks.

“Our original idea was to develop algorithms to automatically improve color maps for data visualization, to make them easier to understand and interpret,” Bujack said. So the team was surprised when they discovered they were the first to determine that the longstanding application of Riemannian geometry, which allows generalizing straight lines to curved surfaces, didn’t work.

To create industry standards, a precise mathematical model of perceived color space is needed. First attempts used Euclidean spaces—the familiar geometry taught in many high schools; more advanced models used Riemannian geometry. The models plot red, green and blue in the 3D space. Those are the colors registered most strongly by light-detecting cones on our retinas, and—not surprisingly—the colors that blend to create all the images on your RGB computer screen.

In the study, which blends psychology, biology and mathematics, Bujack and her colleagues discovered that using Riemannian geometry overestimates the perception of large color differences. That’s because people perceive a big difference in color to be less than the sum you would get if you added up small differences in color that lie between two widely separated shades.

Riemannian geometry cannot account for this effect.

“We didn’t expect this, and we don’t know the exact geometry of this new color space yet,” Bujack said. “We might be able to think of it normally but with an added dampening or weighing function that pulls long distances in, making them shorter. But we can’t prove it yet.”

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

The non-Riemannian nature of perceptual color space by Roxana Bujack, Emily Teti, Jonah Miller, Elektra Caffrey, and Terece L. Turton. Proceedings of the National Academy of Sciences (PNAS) 119 (18) e2119753119 DOI: https://doi.org/10.1073/pnas.2119753119 Published: April 29, 2022

This paper is behind a paywall.

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

Leave a reply
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.

Carbon nanotube optics and the quantum

Leave a reply

A US-France-Germany collaboration has led to some intriguing work with carbon nanotubes. From a June 18, 2018 news item on ScienceDaily,

Researchers at Los Alamos and partners in France and Germany are exploring the enhanced potential of carbon nanotubes as single-photon emitters for quantum information processing. Their analysis of progress in the field is published in this week’s edition of the journal Nature Materials.

“We are particularly interested in advances in nanotube integration into photonic cavities for manipulating and optimizing light-emission properties,” said Stephen Doorn, one of the authors, and a scientist with the Los Alamos National Laboratory site of the Center for Integrated Nanotechnologies (CINT). “In addition, nanotubes integrated into electroluminescent devices can provide greater control over timing of light emission and they can be feasibly integrated into photonic structures. We are highlighting the development and photophysical probing of carbon nanotube defect states as routes to room-temperature single photon emitters at telecom wavelengths.”

A June 18, 2018 Los Alamos National Laboratory (LANL) news release (also on EurekAlert), which originated the news item, expands on the theme,

The team’s overview was produced in collaboration with colleagues in Paris (Christophe Voisin [Ecole Normale Supérieure de Paris (ENS)]) who are advancing the integration of nanotubes into photonic cavities for modifying their emission rates, and at Karlsruhe (Ralph Krupke [Karlsruhe Institute of Technology (KIT]) where they are integrating nanotube-based electroluminescent devices with photonic waveguide structures. The Los Alamos focus is the analysis of nanotube defects for pushing quantum emission to room temperature and telecom wavelengths, he said.

As the paper notes, “With the advent of high-speed information networks, light has become the main worldwide information carrier. . . . Single-photon sources are a key building block for a variety of technologies, in secure quantum communications metrology or quantum computing schemes.”

The use of single-walled carbon nanotubes in this area has been a focus for the Los Alamos CINT team, where they developed the ability to chemically modify the nanotube structure to create deliberate defects, localizing excitons and controlling their release. Next steps, Doorn notes, involve integration of the nanotubes into photonic resonators, to provide increased source brightness and to generate indistinguishable photons. “We need to create single photons that are indistinguishable from one another, and that relies on our ability to functionalize tubes that are well-suited for device integration and to minimize environmental interactions with the defect sites,” he said.

“In addition to defining the state of the art, we wanted to highlight where the challenges are for future progress and lay out some of what may be the most promising future directions for moving forward in this area. Ultimately, we hope to draw more researchers into this field,” Doorn said.

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

Carbon nanotubes as emerging quantum-light sources by X. He, H. Htoon, S. K. Doorn, W. H. P. Pernice, F. Pyatkov, R. Krupke, A. Jeantet, Y. Chassagneux & C. Voisin. Nature Materials (2018) DOI: https://doi.org/10.1038/s41563-018-0109-2 Published online June 18, 2018

This paper is behind a paywall.