PRESS-NEWS.org - Press Release Distribution
PRESS RELEASES DISTRIBUTION

Photonic processor could enable ultrafast AI computations with extreme energy efficiency

This new device uses light to perform the key operations of a deep neural network on a chip, opening the door to high-speed processors that can learn in real-time

2024-12-02
(Press-News.org)

The deep neural network models that power today’s most demanding machine-learning applications have grown so large and complex that they are pushing the limits of traditional electronic computing hardware.

Photonic hardware, which can perform machine-learning computations with light, offers a faster and more energy-efficient alternative. However, there are some types of neural network computations that a photonic device can’t perform, requiring the use of off-chip electronics or other techniques that hamper speed and efficiency.

Building on a decade of research, scientists from MIT and elsewhere have developed a new photonic chip that overcomes these roadblocks. They demonstrated a fully integrated photonic processor that can perform all the key computations of a deep neural network optically on the chip.

The optical device was able to complete the key computations for a machine-learning classification task in less than half a nanosecond while achieving more than 92 percent accuracy — performance that is on par with traditional hardware. 

The chip, composed of interconnected modules that form an optical neural network, is fabricated using commercial foundry processes, which could enable the scaling of the technology and its integration into electronics.

In the long run, the photonic processor could lead to faster and more energy-efficient deep learning for computationally demanding applications like lidar, scientific research in astronomy and particle physics, or high-speed telecommunications.

“There are a lot of cases where how well the model performs isn’t the only thing that matters, but also how fast you can get an answer. Now that we have an end-to-end system that can run a neural network in optics, at a nanosecond time scale, we can start thinking at a higher level about applications and algorithms,” says Saumil Bandyopadhyay ’17, MEng ’18, PhD ’23, a visiting scientist in the Quantum Photonics and AI Group within the Research Laboratory of Electronics (RLE) and a postdoc at NTT Research, Inc., who is the lead author of a paper on the new chip.

Bandyopadhyay is joined on the paper by Alexander Sludds ’18, MEng ’19, PhD ’23, Nicholas Harris PhD ’17, and Darius Bunandar PhD ’19; Stefan Krastanov, a former RLE research scientist who is now an assistant professor at the University of Massachusetts at Amherst; Ryan Hamerly, a visiting scientist at RLE and senior scientist at NTT Research; Matthew Streshinsky, a former silicon photonics lead at Nokia who is now co-founder and CEO of Enosemi; Michael Hochberg, president of Periplous, LLC; and senior author Dirk Englund, a professor in the Department of Electrical Engineering and Computer Science, principal investigator of the Quantum Photonics and Artificial Intelligence Group and of RLE. The research appears today in Nature Photonics.

Machine learning with light

Deep neural networks are composed of many interconnected layers of nodes, or neurons, that operate on input data to produce an output. One key operation in a deep neural network involves the use of linear algebra to perform matrix multiplication, which transforms data as it is passed from layer to layer.

But in addition to these linear operations, deep neural networks perform nonlinear operations that help the model learn more intricate patterns. Nonlinear operations, like activation functions, give deep neural networks the power to solve complex problems. 

In 2017, Englund’s group, along with researchers in the lab of Marin Soljačić, the Cecil and Ida Green Professor of Physics, demonstrated an optical neural network on a single photonic chip that could perform matrix multiplication with light. 

But at the time, the device couldn’t perform nonlinear operations on the chip. Optical data had to be converted into electrical signals and sent to a digital processor to perform nonlinear operations. 

“Nonlinearity in optics is quite challenging because photons don’t interact with each other very easily. That makes it very power consuming to trigger optical nonlinearities, so it becomes challenging to build a system that can do it in a scalable way,” Bandyopadhyay explains.

They overcame that challenge by designing devices called nonlinear optical function units (NOFUs), which combine electronics and optics to implement nonlinear operations on the chip.

The researchers built an optical deep neural network on a photonic chip using three layers of devices that perform linear and nonlinear operations.

A fully-integrated network

At the outset, their system encodes the parameters of a deep neural network into light. Then, an array of programmable beamsplitters, which was demonstrated in the 2017 paper, performs matrix multiplication on those inputs.

The data then pass to programmable NOFUs, which implement nonlinear functions by siphoning off a small amount of light to photodiodes that convert optical signals to electric current. This process, which eliminates the need for an external amplifier, consumes very little energy.

“We stay in the optical domain the whole time, until the end when we want to read out the answer. This enables us to achieve ultra-low latency,” Bandyopadhyay says.

Achieving such low latency enabled them to efficiently train a deep neural network on the chip, a process known as in situ training that typically consumes a huge amount of energy in digital hardware.

“This is especially useful for systems where you are doing in-domain processing of optical signals, like navigation or telecommunications, but also in systems that you want to learn in real time,” he says.

The photonic system achieved more than 96 percent accuracy during training tests and more than 92 percent accuracy during inference, which is comparable to traditional hardware. In addition, the chip performs key computations in less than half a nanosecond.

“This work demonstrates that computing — at its essence, the mapping of inputs to outputs — can be compiled onto new architectures of linear and nonlinear physics that enable a fundamentally different scaling law of computation versus effort needed,” says Englund.

The entire circuit was fabricated using the same infrastructure and foundry processes that produce CMOS computer chips. This could enable the chip to be manufactured at scale, using tried-and-true techniques that introduce very little error into the fabrication process.

Scaling up their device and integrating it with real-world electronics like cameras or telecommunications systems will be a major focus of future work, Bandyopadhyay says. In addition, the researchers want to explore algorithms that can leverage the advantages of optics to train systems faster and with better energy efficiency.

This research was funded, in part, by the National Science Foundation, the Air Force Office of Scientific Research, and NTT Research.

###

Written by Adam Zewe, MIT News

END



ELSE PRESS RELEASES FROM THIS DATE:

Researchers create a new organoid with all key pancreas cells

Researchers create a new organoid with all key pancreas cells
2024-12-02
Researchers from the Organoid group (previously Clevers group) at the Hubrecht Institute have developed a new organoid that mimics the human fetal pancreas, offering a clearer view of its early development. The researchers were able to recreate a complete structure that includes the three key cell types in the pancreas, which previous organoids couldn’t fully mimic. Notably, the team identified a new stem cell that develops into the three cell types. These findings, published in Cell ...

Stimulating hypothalamus restores walking in paralyzed patients

2024-12-02
Researchers at EPFL and Lausanne University Hospital (CHUV), led by professors Grégoire Courtine and Jocelyne Bloch, have achieved a major milestone in the treatment of spinal cord injuries (SCI). By applying deep brain stimulation (DBS) to an unexpected region in the brain—the lateral hypothalamus (LH)—the team has improved the recovery of lower limb movements in two individuals with partial SCI, greatly improving their autonomy and well-being. Wolfgang Jäger, a 54-year-old from Kappel, Austria, has been in a wheelchair ...

Pioneering digital twin model elevates lithium-ion battery performance and safety

Pioneering digital twin model elevates lithium-ion battery performance and safety
2024-12-02
Source: Beijing Institute of Technology Press   Lithium-ion batteries are celebrated for their high specific energy, long service life, and low self-discharge rates. However, ensuring their reliability and safety under various operating conditions is critical for their continued success in industrial applications. Digital twin technology, which creates a virtual replica of a physical entity, offers a promising solution by enabling real-time monitoring and optimization of battery performance. This technology facilitates interactive feedback, data fusion, and iterative ...

A breakthrough in battery capacity degradation analysis and knee point prediction

A breakthrough in battery capacity degradation analysis and knee point prediction
2024-12-02
Analyzing capacity degradation characteristics and accurately predicting the knee point of capacity are crucial for the safety management of lithium-ion batteries (LIBs). A recent breakthrough study presented by researchers from Shandong University introduces a knee point prediction method based on neural network. This advanced method can help us clarify the degradation mechanism and predict the knee point. The study focuses on battery life, which is the result of multiple coupling aging mechanisms affected by multiple factors. It is significantly necessary to clarify the mechanism for the capacity degradation at each stage and possess the ability to detect the knee point. It can not only ...

Newfound mechanism may explain why some cancer treatments boost risk of heart disease

2024-12-02
A cancer therapy that prompts the body’s immune defenses against viruses and bacteria to attack tumors can make patients more vulnerable to heart attack and stroke. A possible explanation for this side effect is that the treatment interferes with immune regulation in the heart’s largest blood vessels, a new study suggests. Led by researchers at NYU Langone Health and its Perlmutter Cancer Center, the new work focused on a potent class of cancer-fighting drugs called immune checkpoint inhibitors. These medications ...

Research alert: How artificial intelligence could automate genomics research

Research alert: How artificial intelligence could automate genomics research
2024-12-02
Researchers at University of California San Diego School of Medicine have demonstrated that large language models (LLMs), such as GPT-4, could help automate functional genomics research, which seeks to determine what genes do and how they interact. The most frequently-used approach in functional genomics, called gene set enrichment, aims to determine the function of experimentally-identified gene sets by comparing them to existing genomics databases. However, more interesting and novel biology is often beyond the scope of established databases. Using artificial intelligence (AI) to analyze gene ...

‘I don’t feel your pain’: How alcohol increases aggression

2024-12-02
COLUMBUS, Ohio – Alcohol’s ability to increase people’s pain threshold is one reason that drinking also leads to more aggressive behavior, a new study suggests.   Researchers found that the less pain that study participants felt after drinking an alcoholic beverage, the more pain they were willing to inflict on someone else.   “We’ve all heard the idiom ‘I feel your pain,’” said study co-author Brad Bushman, professor of communication at The Ohio State University.   “But if intoxicated people can’t feel their own pain, they might be less likely to feel ...

The Microprocessor inside you

The Microprocessor inside you
2024-12-02
It’s a big year for microRNAs. The 2024 Nobel Prize in Physiology or Medicine went to Victor Ambros and Gary Ruvkun, who discovered the first microRNA in 1993. Today, we know that humans make more than 1,000 different microRNAS. These molecules are critical for building and maintaining healthy bodies, so it’s crucial that they’re made the right way. Errors in microRNA manufacture can put us at risk for developmental disorders, cancer, or neurodegenerative disease. To learn how cells accurately generate a mind-boggling array of microRNAs, Cold Spring Harbor Laboratory (CSHL) Professor and HHMI Investigator Leemor ...

Landmark World Drought Atlas reveals systemic nature of hazard risks, underlines need for national plans, international cooperation

Landmark World Drought Atlas reveals systemic nature of hazard risks, underlines need for national plans, international cooperation
2024-12-02
Riyadh, Saudi Arabia — As record-breaking droughts are becoming a new normal around the globe, the UN Convention to Combat Desertification (UNCCD) and the European Commission Joint Research Centre (JRC) launch the most comprehensive global publication on drought risks and solutions as an urgent wake-up call for world leaders and citizens. The landmark new World Drought Atlas depicts the systemic nature of drought risks through dozens of maps, infographics, and case studies. It illustrates how drought risks are interconnected across sectors like energy, agriculture, river transport, and international trade and how they can trigger cascading effects, fueling inequalities and ...

To build better fiber optic cables, ask a clam

To build better fiber optic cables, ask a clam
2024-12-02
DURHAM, N.C. -- Since the first fiber optic cables rolled out in the 1970s, they’ve become a major part of everything from medical devices to high-speed internet and cable TV. But as it turns out, one group of marine mollusks was way ahead of us. A new study reveals that clams called heart cockles -– so-named because of their heart-shaped shells -- have unique structures in their shells that act like fiber optic cables to convey specific wavelengths of light into the bivalves’ tissues. Researchers from Duke University and Stanford University used electron and laser microscopy and computer simulations ...

LAST 30 PRESS RELEASES:

‘Altar tent’ discovery puts Islamic art at the heart of medieval Christianity

Policy briefs present approach for understanding prison violence

Early adult mortality is higher than expected in US post-COVID

Recycling lithium-ion batteries cuts emissions and strengthens supply chain

Study offers new hope for relieving chronic pain in dialysis patients

How does the atmosphere affect ocean weather?

Robots get smarter to work in sewers

Speech Accessibility Project data leads to recognition improvements on Microsoft Azure

Tigers in the neighborhood: How India makes room for both tigers and people

Grove School’s Arthur Paul Pedersen publishes critical essay on scientific measurement literacy

Moffitt study finds key biomarker to predict KRASG12C inhibitor effectiveness in lung cancer

Improving blood transfusion monitoring in critical care patients: Insights from diffuse optics

Powerful legal and financial services enable kleptocracy, research shows

Carbon capture from constructed wetlands declines as they age

UCLA-led study establishes link between early side effects from prostate cancer radiation and long-term side effects

Life cycles of some insects adapt well to a changing climate. Others, not so much.

With generative AI, MIT chemists quickly calculate 3D genomic structures

The gut-brain connection in Alzheimer’s unveiled with X-rays

NIH-funded clinical trial will evaluate new dengue therapeutic

Sound is a primary issue in the lives of skateboarders, study shows

Watch what you eat: NFL game advertisements promote foods high in fat, sodium

Red Dress Collection Concert hosted by Sharon Stone kicks off American Heart Month

One of the largest studies on preterm birth finds a maternal biomarker test significantly reduces neonatal morbidities and improves neonatal outcomes

One of the largest studies of its kind finds early intervention with iron delivered intravenously during pregnancy is a safe and effective treatment for anemia

New Case Western Reserve University study identifies key protein’s role in psoriasis

First-ever ethics checklist for portable MRI brain researchers

Addressing 3D effects of clouds for significant improvements of climate models

Gut microbes may mediate the link between drinking sugary beverages and diabetes risk

Ribosomes team up in difficult situations, new technology shows

Mortality trends among adults ages 25-44 in the US

[Press-News.org] Photonic processor could enable ultrafast AI computations with extreme energy efficiency
This new device uses light to perform the key operations of a deep neural network on a chip, opening the door to high-speed processors that can learn in real-time