Connect with us

Bioengineer

Researchers present spontaneous sparse learning for PCM-based memristor neural networks

Credit: UNIST An international team of researchers, affiliated with UNIST has unveiled a novel technology that could improve the learning

Published

on

An international team of researchers, affiliated with UNIST has unveiled a novel technology that could improve the learning ability of artificial neural networks (ANNs).

Professor Hongsik Jeong and his research team in the Department of Materials Science and Engineering at UNIST, in collaboration with researchers from Tsinghua University in China, proposed a new learning method to improve the learning ability of ANN chips by challenging its instability.

Artificial neural network chips are capable of mimicking the structural, functional and biological features of human neural networks, and thus have been considered the technology of the future. In this study, the research team demonstrated the effectiveness of the proposed learning method by building phase change memory (PCM) memristor arrays that operate like ANNs. This learning method is also advantageous in that its learning ability can be improved without additional power consumption, since PCM undergoes a spontaneous resistance increase due to the structural relaxation after amorphization.

ANNs, like human brains, use less energy even when performing computation and memory tasks, simultaneously. However, the artificial neural network chip in which a large number of physical devices are integrated has a disadvantage that there is an error. The existing artificial neural network learning method assumes a perfect artificial neural network chip with no errors, so the learning ability of the artificial neural network is poor.

The research team developed a memristor artificial neural network learning method based on a phase-change memory, conceiving that the real human brain does not require near-perfect motion. This learning method reflects the “resistance drift” (increased electrical resistance) of the phase change material in the memory semiconductor in learning. During the learning process, since the information update pattern is recorded in the form of increasing electrical resistance in the memristor, which serves as a synapse, the synapse additionally learns the association between the pattern it changes and the data it is learning.

The research team showed that the learning method developed through an experiment to classify handwriting composed of numbers 0-9 has an effect of improving learning ability by about 3%. In particular, the accuracy of the number 8, which is difficult to classify handwriting, has improved significantly. The learning ability improved thanks to the synaptic update pattern that changes differently according to the difficulty of handwriting classification.

Researchers expect that their findings are expected to promote the learning algorithms with the intrinsic properties of memristor devices, opening a new direction for development of neuromorphic computing chips.

###

The findings of this research have been published in the online version of Nature Communications on January 12, 2021. This study has been supported by the National Research Foundation of Korea (NRF), UNIST’s Future Leading Specialized Research Project, and the UNIST Graduate School of Artificial Intelligence.

Researchers Present Spontaneous Sparse Learning for PCM-based Memristor Neural Networks

The research team developed a memristor artificial neural network learning method based on a phase-change memory, conceiving that the real human brain does not require near-perfect motion. This learning method reflects the “resistance drift” (increased electrical resistance) of the phase change material in the memory semiconductor in learning. During the learning process, since the information update pattern is recorded in the form of increasing electrical resistance in the memristor, which serves as a synapse, the synapse additionally learns the association between the pattern it changes and the data it is learning.

Source: https://bioengineer.org/researchers-present-spontaneous-sparse-learning-for-pcm-based-memristor-neural-networks/

researchers-present-spontaneous-sparse-learning-for-pcm-based-memristor-neural-networks

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Bioengineer

NYU Abu Dhabi researchers design simulator to help stop the spread of ‘fake news’

The new game, Fakey, emulates a social media feed and teaches users to recognize credible contentCredit: Courtesy of NYU Abu

Published

on

The new game, Fakey, emulates a social media feed and teaches users to recognize credible content

Abu Dhabi, UAE, April 27, 2021: As people around the world increasingly get their news from social media, online misinformation has emerged as an area of great concern. To improve news literacy and reduce the spread of misinformation, NYUAD Center for Cybersecurity researcher and lead author Nicholas Micallef is part of a team that designed Fakey, a game that emulates a social media news feed and prompts players to use available signals to recognize and scrutinize suspicious content and focus on credible information. Players can share, like, or fact-check individual articles.

In a new study, Fakey: A Game Intervention to Improve News Literacy on Social Media published in the ACM Digital Library, Micallef and his colleagues Mihai Avram, Filippo Menczer, and Sameer Patil from the Luddy School of Informatics, Computing, and Engineering, Indiana University, present the analysis of interactions with Fakey, which was released to the general public as a web and mobile app with data procured after 19 months of use. Interviews were conducted to verify player understanding of the game elements. The researchers found that the more players interacted with articles in the game, the better their skills at spotting credible content became. However, playing the game did not affect players’ ability to recognize questionable content. Further research will help determine how much gameplay would be necessary to be able to distinguish between legitimate and questionable content.

Games like Fakey, which was designed and developed by researchers at Indiana University, could be offered as a tool to social media users. For example, social media platforms could conduct regular exercises (akin to ‘phishing drills’ used in organizations for employee security training) wherein users practice identifying questionable articles. Or, the researchers say, such games could be integrated into media literacy curricula in schools. “The impact of misinformation could be substantially reduced if people were given tools to help them recognize and ignore such content,” said Micallef. “The principles and mechanisms used by Fakey can inform the design of social media functionality in a way that empowers people to distinguish between credible and fake content in their news feeds and increase their digital literacy.”

###

About NYU Abu Dhabi

NYU Abu Dhabi is the first comprehensive liberal arts and science campus in the Middle East to be operated abroad by a major American research university. NYU Abu Dhabi has integrated a highly-selective liberal arts, engineering, and science curriculum with a world center for advanced research and scholarship enabling its students to succeed in an increasingly interdependent world and advance cooperation and progress on humanity’s shared challenges. NYU Abu Dhabi’s high-achieving students have come from more than 115 nations and speak over 115 languages. Together, NYU’s campuses in New York, Abu Dhabi, and Shanghai form the backbone of a unique global university, giving faculty and students opportunities to experience varied learning environments and immersion in other cultures at one or more of the numerous study-abroad sites NYU maintains on six continents.

###

Source: https://bioengineer.org/nyu-abu-dhabi-researchers-design-simulator-to-help-stop-the-spread-of-fake-news/

nyu-abu-dhabi-researchers-design-simulator-to-help-stop-the-spread-of-‘fake-news’

Continue Reading

Bioengineer

The science of sound, vibration to better diagnose, treat brain diseases

Multidisciplinary researchers uncover new ways to use ultrasound energy to image and treat hard-to-reach areas of brainCredit: Allison Carter, Georgia

Published

on

Multidisciplinary researchers uncover new ways to use ultrasound energy to image and treat hard-to-reach areas of brain

A team of engineering researchers at the Georgia Institute of Technology hopes to uncover new ways to diagnose and treat brain ailments, from tumors and stroke to Parkinson’s disease, leveraging vibrations and ultrasound waves.

The five-year, $2 million National Science Foundation (NSF) project initiated in 2019 already has resulted in several published journal articles that offer promising new methods to focus ultrasound waves through the skull, which could lead to broader use of ultrasound imaging — considered safer and less expensive than magnetic resonance imaging (MRI) technology.

Specifically, the team is researching a broad range of frequencies, spanning low frequency vibrations (audio frequency range) and moderate frequency guided waves (100 kHz to 1 MHz) to high frequencies employed in brain imaging and therapy (in the MHz range).

“We’re coming up with a unique framework that incorporates different research perspectives to address how you use sound and vibration to treat and diagnose brain diseases,” explained Costas Arvanitis, an assistant professor in Georgia Tech’s George W. Woodruff School of Mechanical Engineering and the Wallace H. Coulter Department of Biomedical Engineering at Georgia Tech and Emory University. “Each researcher is bringing their own expertise to explore how vibrations and waves across a range of frequencies could either extract information from the brain or focus energy on the brain.”

Accessing the Brain Is a Tough Challenge

While it is possible to treat some tumors and other brain diseases non-invasively if they are near the center of the brain, many other conditions are harder to access, the researchers say.

“The center part of the brain is most accessible; however, even if you are able to target the part of the brain away from the center, you still have to go through the skull,” Arvanitis said.

He added that moving just 1 millimeter in the brain constitutes “a huge distance” from a diagnostic perspective. The science community widely acknowledges the brain’s complexity, each part associated with a different function and brain cells differing from one to the other.

According to Brooks Lindsey, a biomedical engineering assistant professor at Georgia Tech and Emory, there is a reason why brain imaging or therapy works well in some people but not in others.

“It depends on the individual patient’s skull characteristics,” he said, noting that some people have slightly more trabecular bone ? the spongy, porous part of the bone ? that makes it more difficult to treat.

Using ultrasound waves, the researchers are tackling the challenge on multiple levels. Lindsey’s lab uses ultrasound imaging to assess skull properties for effective imaging and therapy. He said his team conducted the first investigation that uses ultrasound imaging to measure the effects of bone microstructure — specifically, the degree of porosity in the inner, trabecular bone layer of the skull.

“By understanding transmission of acoustic waves through microstructure in an individual’s skull, non-invasive ultrasound imaging of the brain and delivery of therapy could be possible in a greater number of people,” he said, explaining one potential application would be to image blood flow in the brain following a stroke.

Refocusing Ultrasound Beams on the Fly

Arvanitis’ lab recently found a new way to focus ultrasound through the skull and into the brain, which is “100-fold faster than any other method,” Arvanitis said. His team’s work in adaptive focusing techniques would allow clinicians to adjust the ultrasound on the fly to focus it better.

“Current systems rely a lot on MRIs, which are big, bulky, and extremely expensive,” he said. “This method lets you adapt and refocus the beam. In the future this could allow us to design less costly, simpler systems, which would make the technology available to a wider population, as well as be able to treat different parts of the brain.”

Using ‘Guided Waves’ to Access Periphery Brain Areas

Another research cohort, led by Alper Erturk, Woodruff Professor of Mechanical Engineering at Georgia Tech, and former Georgia Tech colleague Massimo Ruzzene, Slade Professor of Mechanical Engineering at the University of Colorado Boulder, performs high-fidelity modeling of skull bone mechanics along with vibration-based elastic parameter identification. They also leverage guided ultrasonic waves in the skull to expand the treatment envelope in the brain. Erturk and Ruzzene are mechanical engineers by background, which makes their exploration of vibrations and guided waves in difficult-to-reach brain areas especially fascinating.

Erturk noted that guided waves are used in other applications such as aerospace and civil structures for damage detection. “Accurate modeling of the complex bone geometry and microstructure, combined with rigorous experiments for parameter identification, is crucial for a fundamental understanding to expand the accessible region of the brain,” he said.

Ruzzene compared the brain and skull to the Earth’s core and crust, with the cranial guided waves acting as an earthquake. Just as geophysicists use earthquake data on the Earth’s surface to understand the Earth’s core, so are Erturk and Ruzzene using the guided waves to generate tiny, high frequency “earthquakes” on the external surface of the skull to characterize what comprises the cranial bone.

Trying to access the brain periphery via conventional ultrasound poses added risks from the skull heating up. Fortunately, advances such as cranial leaky Lamb waves increasingly are recognized for transmitting wave energy to that region of the brain.

These cranial guided waves could complement focused ultrasound applications to monitor changes in the cranial bone marrow from health disorders, or to efficiently transmit acoustic signals through the skull barrier, which could help access metastases and treat neurological conditions in currently inaccessible regions of the brain.

Ultimately, the four researchers hope their work will make full brain imaging feasible while stimulating new medical imaging and therapy techniques. In addition to transforming diagnosis and treatment of brain diseases, the techniques could better detect traumas and skull-related defects, map the brain function, and enable neurostimulation. Researchers also see the potential for uncovering ultrasound-based blood-brain barrier openings for drug delivery for managing and treating diseases such as Alzheimer’s.

###

With this comprehensive research of the skull-brain system, and by understanding the fundamentals of transcranial ultrasound, the team hopes to make it more available to even more diseases and target many parts of the brain.

This work is funded by the National Science Foundation (CMMI Award 1933158 “Coupling Skull-Brain Vibroacoustics and Ultrasound Toward Enhanced Imaging, Diagnosis, and Therapy”).

CITATIONS: C. Sugino, M. Ruzzene, and A. Erturk, “Experimental and Computational Investigation of Guided Waves in a Human Skull.” (Ultrasound in Medicine and Biology, 2021) https://doi.org/10.1016/j.ultrasmedbio.2020.11.019

M. Mazzotti, E. Kohtanen, A. Erturk, and M. Ruzzene, “Radiation Characteristics of Cranial Leaky Lamb Waves.” (IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 2021) https://doi.org/10.1109/TUFFC.2021.3057309

S. Schoen, C. Arvanitis, “Heterogeneous Angular Spectrum Method for Trans-Skull Imaging and Focusing.” (IEEE Xplore, 2020) https://ieeexplore.ieee.org/document/8902167

B. Jing, C. Arvanitis, B. Lindsey, “Effect of Incidence Angle and Wave Mode Conversion on Transcranial Ultrafast Doppler Imaging.” (IEEE Xplore, 2020) https://ieeexplore.ieee.org/document/9251477

The Georgia Institute of Technology, or Georgia Tech, is a top 10 public research university developing leaders who advance technology and improve the human condition.

The Institute offers business, computing, design, engineering, liberal arts, and sciences degrees. Its nearly 40,000 students, representing 50 states and 149 countries, study at the main campus in Atlanta, at campuses in France and China, and through distance and online learning.

As a leading technological university, Georgia Tech is an engine of economic development for Georgia, the Southeast, and the nation, conducting more than $1 billion in research annually for government, industry, and society.

https://rh.gatech.edu/news/646931/science-sound-vibration-better-diagnose-treat-brain-diseases

Source: https://bioengineer.org/the-science-of-sound-vibration-to-better-diagnose-treat-brain-diseases/

the-science-of-sound,-vibration-to-better-diagnose,-treat-brain-diseases

Continue Reading

Bioengineer

Nontoxic, flexible energy converters could power wearable devices

Nontoxic, nanotube-based thermoelectric generation converts uneven heat distribution from wearables to electrical energy for their next cycle of operation.Credit: Injung

Published

on

Nontoxic, nanotube-based thermoelectric generation converts uneven heat distribution from wearables to electrical energy for their next cycle of operation.

WASHINGTON, April 27, 2021 — A wide variety of portable and wearable electronics have become a large part of our daily lives, so a group of Stanford University researchers wondered if these could be powered by harvesting electricity from the waste heat that exists all around us.

Further inspiration came from a desire to ultimately fabricate energy converting devices from the same materials as the active devices themselves, so they can blend in as an integral part of the total system. Today, many biomedical nanodevices’ power supplies come from several types of batteries that must be separated from the active portion of the systems, which is not ideal.

In Applied Physics Letters, from AIP Publishing, the researchers report the design and fabrication of single-wall carbon nanotube thermoelectric devices on flexible polyimide substrates as a basis for wearable energy converters.

“Carbon nanotubes are one-dimensional materials, known for good thermoelectric properties, which mean developing a voltage across them in a temperature gradient,” said Eric Pop, a professor of electrical engineering and materials science. “The challenge is that carbon nanotubes also have high thermal conductivity, meaning it’s difficult to maintain a thermal gradient across them, and they have been hard to assemble them into thermoelectric generators at low cost.”

The group uses printed carbon nanotube networks to tackle both challenges.

“For example, carbon nanotube spaghetti networks have much lower thermal conductivity than carbon nanotubes taken alone, due to the presence of junctions in the networks, which block heat flow,” Pop said. “Also, direct printing such carbon nanotube networks can significantly reduce their cost when they are scaled up.”

Thermoelectric devices generate electric power locally “by reusing waste heat from personal devices, appliances, vehicles, commercial and industrial processes, computer servers, time-varying solar illumination, and even the human body,” said Hye Ryoung Lee, lead author and a research scientist.

“To eliminate hindrances to large-scale application of thermoelectric materials — toxicity, materials scarcity, mechanical brittleness — carbon nanotubes offer an excellent alternative to other commonly used materials,” Lee said.

The group’s approach demonstrates a path to using carbon nanotubes with printable electrodes on flexible polymer substrates in a process anticipated to be economical for large-volume manufacturing. It is also “greener” than other processes, because water is used as the solvent and additional dopants are avoided.

Flexible and wearable energy harvesters can be embedded into fabrics or clothes or placed on unusual shapes and form factors.

“In contrast, traditional thermoelectrics that rely on bismuth telluride are brittle and stiff, with limited applications,” Pop said. “Carbon-based thermoelectrics are also more environmentally friendly than those based on rare or toxic materials like bismuth and tellurium.”

The most important concept in the group’s work is to “recycle energy as much as we can, converting uneven heat distribution to electrical energy for use for the next cycle of operation, which we demonstrated by using nontoxic nanotube-based thermoelectric generation,” said Yoshio Nishi, a professor of electrical engineering. “This concept is in full alliance with the world’s goal of reducing our total energy consumption.”

###

The article “Carbon nanotube thermoelectric devices by direct printing: Towards wearable energy converters” is authored by Hye Ryoung Lee, Naoki Furukawa, Antonio J. Ricco, Eric Pop, Yi Cui, and Yoshio Nishi. The article will appear in Applied Physics Letters on April 27 (DOI: 10.1063/5.0042349). After that date, it can be accessed at https://aip.scitation.org/doi/10.1063/5.0042349.

ABOUT THE JOURNAL

Applied Physics Letters features rapid reports on significant discoveries in applied physics. The journal covers new experimental and theoretical research on applications of physics phenomena related to all branches of science, engineering, and modern technology. See https://aip.scitation.org/journal/apl.

In Applied Physics Letters, from AIP Publishing, the researchers report the design and fabrication of single-wall carbon nanotube thermoelectric devices on flexible polyimide substrates as a basis for wearable energy converters.

Source: https://bioengineer.org/nontoxic-flexible-energy-converters-could-power-wearable-devices/

nontoxic,-flexible-energy-converters-could-power-wearable-devices

Continue Reading

Title

Reuters2 hours ago

Disneyland Paris to re-open on June 17

Disneyland Paris (DIS.N) said on Monday that it would re-open on June 17, as French bars, restaurants and tourism sites...

ZDNET7 hours ago

How Crocs used robots to rule the comfort economy

Sweatpants and comfortable kicks have had a heck of a run during the pandemic. You can thank the robots.

CNBC10 hours ago

Airbnb says first-quarter revenue rose 5% as vacationers return to travel

Airbnb's net loss tripled, but the company expects its adjusted margin to improve in the second half of the year...

Business insider13 hours ago

SES Government Solutions Provides Medium Earth Orbit Satellite Services for Combatant Command

SES Government Solutions, a wholly-owned subsidiary of SES, in close partnership with a key U.S. Government customer, designed, developed and...

Cointelegraph15 hours ago

Here’s how Bitcoin’s intraday volatility complicates leverage trading

Derivatives exchanges offer up to 100x leverage, but traders must consider how Bitcoin's intraday volatility increases their liquidation risk.

Crunchbase1 day ago

Exclusive: Forager Chews On $4M To Digitize Local Food Access

Its platform digitizes and streamlines the discovery of new local food vendors, onboarding and management of those relationships.

Blockchain news1 day ago

Internet Computer (ICP): Everything You Need to Know

After Internet Computer ICP tokens were listed on a number of leading cryptocurrency exchanges, its price even exceeded the maximum...

Ventureburn1 day ago

AlphaCode awards R2-million and support to fintech startups

The 10 startups, which have just completed a three-month programme, competed for one of four places in an extended 6-month...

Entrepreneur1 day ago

7 Quick Ways to Make Money Investing $1,000

If you're shrewd, you can turn one thousand bucks into even more money. Here's how.

Bioengineer2 days ago

NYU Abu Dhabi researchers design simulator to help stop the spread of ‘fake news’

The new game, Fakey, emulates a social media feed and teaches users to recognize credible contentCredit: Courtesy of NYU Abu

Review

    Select language

    Trending