GE
-
KAIST Secures Core Technology for Ultra-High-Resolution Image Sensors
A joint research team from Korea and the United States has developed next-generation, high-resolution image sensor technology with higher power efficiency and a smaller size compared to existing sensors. Notably, they have secured foundational technology for ultra-high-resolution shortwave infrared (SWIR) image sensors, an area currently dominated by Sony, paving the way for future market entry.
KAIST (represented by President Kwang Hyung Lee) announced on the 20th of November that a research team led by Professor SangHyeon Kim from the School of Electrical Engineering, in collaboration with Inha University and Yale University in the U.S., has developed an ultra-thin broadband photodiode (PD), marking a significant breakthrough in high-performance image sensor technology.
This research drastically improves the trade-off between the absorption layer thickness and quantum efficiency found in conventional photodiode technology. Specifically, it achieved high quantum efficiency of over 70% even in an absorption layer thinner than one micrometer (μm), reducing the thickness of the absorption layer by approximately 70% compared to existing technologies.
A thinner absorption layer simplifies pixel processing, allowing for higher resolution and smoother carrier diffusion, which is advantageous for light carrier acquisition while also reducing the cost. However, a fundamental issue with thinner absorption layers is the reduced absorption of long-wavelength light.
< Figure 1. Schematic diagram of the InGaAs photodiode image sensor integrated on the Guided-Mode Resonance (GMR) structure proposed in this study (left), a photograph of the fabricated wafer, and a scanning electron microscope (SEM) image of the periodic patterns (right) >
The research team introduced a guided-mode resonance (GMR) structure* that enables high-efficiency light absorption across a wide spectral range from 400 nanometers (nm) to 1,700 nanometers (nm). This wavelength range includes not only visible light but also light the SWIR region, making it valuable for various industrial applications.
*Guided-Mode Resonance (GMR) Structure: A concept used in electromagnetics, a phenomenon in which a specific (light) wave resonates (forming a strong electric/magnetic field) at a specific wavelength. Since energy is maximized under these conditions, it has been used to increase antenna or radar efficiency.
The improved performance in the SWIR region is expected to play a significant role in developing next-generation image sensors with increasingly high resolutions. The GMR structure, in particular, holds potential for further enhancing resolution and other performance metrics through hybrid integration and monolithic 3D integration with complementary metal-oxide-semiconductor (CMOS)-based readout integrated circuits (ROIC).
< Figure 2. Benchmark for state-of-the-art InGaAs-based SWIR pixels with simulated EQE lines as a function of TAL variation. Performance is maintained while reducing the absorption layer thickness from 2.1 micrometers or more to 1 micrometer or less while reducing it by 50% to 70% >
The research team has significantly enhanced international competitiveness in low-power devices and ultra-high-resolution imaging technology, opening up possibilities for applications in digital cameras, security systems, medical and industrial image sensors, as well as future ultra-high-resolution sensors for autonomous driving, aerospace, and satellite observation.
Professor Sang Hyun Kim, the lead researcher, commented, “This research demonstrates that significantly higher performance than existing technologies can be achieved even with ultra-thin absorption layers.”
< Figure 3. Top optical microscope image and cross-sectional scanning electron microscope image of the InGaAs photodiode image sensor fabricated on the GMR structure (left). Improved quantum efficiency performance of the ultra-thin image sensor (red) fabricated with the technology proposed in this study (right) >
The results of this research were published on 15th of November, in the prestigious international journal Light: Science & Applications (JCR 2.9%, IF=20.6), with Professor Dae-Myung Geum of Inha University (formerly a KAIST postdoctoral researcher) and Dr. Jinha Lim (currently a postdoctoral researcher at Yale University) as co-first authors. (Paper title: “Highly-efficient (>70%) and Wide-spectral (400 nm -1700 nm) sub-micron-thick InGaAs photodiodes for future high-resolution image sensors”)
This study was supported by the National Research Foundation of Korea.
2024.11.22 View 217 -
KAIST Proposes AI Training Method that will Drastically Shorten Time for Complex Quantum Mechanical Calculations
- Professor Yong-Hoon Kim's team from the School of Electrical Engineering succeeded for the first time in accelerating quantum mechanical electronic structure calculations using a convolutional neural network (CNN) model
- Presenting an AI learning principle of quantum mechanical 3D chemical bonding information, the work is expected to accelerate the computer-assisted designing of next-generation materials and devices
The close relationship between AI and high-performance scientific computing can be seen in the fact that both the 2024 Nobel Prizes in Physics and Chemistry were awarded to scientists for their AI-related research contributions in their respective fields of study. KAIST researchers succeeded in dramatically reducing the computation time for highly sophisticated quantum mechanical computer simulations by predicting atomic-level chemical bonding information distributed in 3D space using a novel AI approach.
KAIST (President Kwang-Hyung Lee) announced on the 30th of October that Professor Yong-Hoon Kim's team from the School of Electrical Engineering developed a 3D computer vision artificial neural network-based computation methodology that bypasses the complex algorithms required for atomic-level quantum mechanical calculations traditionally performed using supercomputers to derive the properties of materials.
< Figure 1. Various methodologies are utilized in the simulation of materials and materials, such as quantum mechanical calculations at the nanometer (nm) level, classical mechanical force fields at the scale of tens to hundreds of nanometers, continuum dynamics calculations at the macroscopic scale, and calculations that mix simulations at different scales. These simulations are already playing a key role in a wide range of basic research and application development fields in combination with informatics techniques. Recently, there have been active efforts to introduce machine learning techniques to radically accelerate simulations, but research on introducing machine learning techniques to quantum mechanical electronic structure calculations, which form the basis of high-scale simulations, is still insufficient. >
The quantum mechanical density functional theory (DFT) calculations using supercomputers have become an essential and standard tool in a wide range of research and development fields, including advanced materials and drug design, as they allow fast and accurate prediction of material properties.
*Density functional theory (DFT): A representative theory of ab initio (first principles) calculations that calculate quantum mechanical properties from the atomic level.
However, practical DFT calculations require generating 3D electron density and solving quantum mechanical equations through a complex, iterative self-consistent field (SCF)* process that must be repeated tens to hundreds of times. This restricts its application to systems with only a few hundred to a few thousand atoms.
*Self-consistent field (SCF): A scientific computing method widely used to solve complex many-body problems that must be described by a number of interconnected simultaneous differential equations.
Professor Yong-Hoon Kim’s research team questioned whether recent advancements in AI techniques could be used to bypass the SCF process. As a result, they developed the DeepSCF model, which accelerates calculations by learning chemical bonding information distributed in a 3D space using neural network algorithms from the field of computer vision.
< Figure 2. The deepSCF methodology developed in this study provides a way to rapidly accelerate DFT calculations by avoiding the self-consistent field process (orange box) that had to be performed repeatedly in traditional quantum mechanical electronic structure calculations through artificial neural network techniques (green box). The self-consistent field process is a process of predicting the 3D electron density, constructing the corresponding potential, and then solving the quantum mechanical Cohn-Sham equations, repeating tens to hundreds of times. The core idea of the deepSCF methodology is that the residual electron density (δρ), which is the difference between the electron density (ρ) and the sum of the electron densities of the constituent atoms (ρ0), corresponds to chemical bonding information, so the self-consistent field process is replaced with a 3D convolutional neural network model. >
The research team focused on the fact that, according to density functional theory, electron density contains all quantum mechanical information of electrons, and that the residual electron density — the difference between the total electron density and the sum of the electron densities of the constituent atoms — contains chemical bonding information. They used this as the target for machine learning.
They then adopted a dataset of organic molecules with various chemical bonding characteristics, and applied random rotations and deformations to the atomic structures of these molecules to further enhance the model’s accuracy and generalization capabilities. Ultimately, the research team demonstrated the validity and efficiency of the DeepSCF methodology on large, complex systems.
< Figure 3. An example of applying the deepSCF methodology to a carbon nanotube-based DNA sequence analysis device model (top left). In addition to classical mechanical interatomic forces (bottom right), the residual electron density (top right) and quantum mechanical electronic structure properties such as the electronic density of states (DOS) (bottom left) containing information on chemical bonding are rapidly predicted with an accuracy corresponding to the standard DFT calculation results that perform the SCF process. >
Professor Yong-Hoon Kim, who supervised the research, explained that his team had found a way to map quantum mechanical chemical bonding information in a 3D space onto artificial neural networks. He noted, “Since quantum mechanical electron structure calculations underpin materials simulations across all scales, this research establishes a foundational principle for accelerating material calculations using artificial intelligence.”
Ryong-Gyu Lee, a PhD candidate in the School of Electrical Engineering, served as the first author of this research, which was published online on October 24 in Npj Computational Materials, a prestigious journal in the field of material computation. (Paper title: “Convolutional network learning of self-consistent electron density via grid-projected atomic fingerprints”)
This research was conducted with support from the KAIST High-Risk Research Program for Graduate Students and the National Research Foundation of Korea’s Mid-career Researcher Support Program.
2024.10.30 View 910 -
Team KAIST Crowned Champion for their World’s Best Ironman, Winning their 2nd Consecutive Win at the Cyborg Olympics
< Group photo of our research team participating in the 3rd Cybathlon >
The Cybathlon is an international competition, which aims to help overcome disabilities through the use of robotics technology. KAIST researchers came in third, winning the bronze medal, at their 1st participation in 2016, won their first gold medal at the 2nd competition in 2020, and took back the gold medal at this year’s 3rd competition, successfully defending their champion title.
KAIST (President Kwang-Hyung Lee) announced on the 28th of October that the wearable robot ‘WalkON Suit F1’ developed by the joint research team of KAIST EXO-Lab, Move Lab, and Angel Robotics, led by Professor Kyoungchul Kong of the Department of Mechanical Engineering (Chairman of Angel Robotics), won gold medal in Cybathlon 2024 held on the 27th.
< Seunghwan Kim, the Paraplegic Pilot of Team KAIST, cheering after completing all the missions >
Cybathlon is an international competition first held in Switzerland that challenges teams to develop assistive technologies with the purpose of overcoming disabilities, and is also called the Cyborg Olympics. Right after each competition, the missions for the next competition are announced, and research teams from around the world research and develop the right assistive technology for next four years to accomplish the given mission.
Aside from the Exoskeleton Race, competitions in a total of eight disciplines are held, including Arm Prosthesis Race, Leg Prosthesis Race, and Wheelchair Race. A total of 71 teams from 26 countries participated in the 3rd Cybathlon event. Professor Kyoungchul Kong’s research team participated in the Exoskeleton Race, the same discipline they took part in the previous competition.
The Exoskeleton Race is the highlight of the event that can be called the core of the Cybathlon. In the prosthetic arm or leg events, disabled athletes wearing traditional assistive devices instead of powered devices often win depending more on the ability of the athletes rather than the technology behind the tools.
However, the exoskeleton event requires individuals with paraplegic disability to walk completely dependent on the robotic device to perform various missions, so the technical difficulty and the dependence on robotics technology is high.
In fact, many teams gave up on participating after seeing the missions for this competition, and more than half of the research teams declared withdrawal during the developmental process. In the end, only six teams from Korea, Thailands, Switzerland, Germany, and the Netherlands participated in the actual competition. Even the research team from the Swiss headquarters declared forfeiture midway as the competition date drew nearer.
< Cybathlon 2024 – Exoskeleton Race Mission Description >
The reason why many teams gave up in the exoskeleton discipline was because the difficulty of the missions was unusually high. Most research teams have the skills to make paraplegic athletes walk, but there were many other difficult tasks, such as making them walk without crutches or using both hands, while standing free on both legs on the exoskeleton, to cut a lump of sponge block, as in imitating food preparing process.
The reason why the difficulty of the missions increased like this is because Professor Kong's research team completed all the missions given to them too quickly in the last competition. In fact, in the last competition, there was even a question asked whether Kim Byeong-Uk (paraplegic) wearing the WalkON Suit F1 was really disabled.
Professor Kong's research team developed WalkON Suit F1 to successfully complete the missions. The number of motorized joints increased from six to twelve, and the output of the motor itself was more than doubled compared to the previous model that ran in the competition back in 2020. The 6-channel ground reaction force sensor on the foot measured the robot's balance 1,000 times per second to maintain balance. Cameras were installed to detect obstacles, and an AI board for implementation of AI neural network was also installed.
On top of the technologies required to complete the competition missions, a function was implemented that allows the users to wear the robot by themselves and dock on to it right from their wheelchair. In the process, all parts were domestically produced and all basic technologies were internalized. The outer design of the robot was done by Professor Hyunjoon Park of KAIST, and the harmony between people and robots was pursued.
In the end, the results of the competition were as expected. The only team that could successfully perform all of the mission tasks, which were originally designed to challenge Professor Kong's research team, was Professor Kong's team. They successfully completed missions such as moving by sidesteps between narrow chairs, moving boxes, walking freely unassisted by crutches, passing through a narrow door and closing it behind, and working on food preparation in the kitchen, recording 6 minutes and 41 seconds to complete all six tasks.
The Swiss and Thai teams that took 2nd and 3rd places were all given 10 minutes, but only were able to perform two missions, each earning twenty points. It was an unevenly matched race to begin with. The Cybathlon broadcast team was more surprised and interested in the performance of WalkON Suit F1 than in result of the race.
< Team KAIST’s Paraplegic Pilot Seunghwan Kim (left), and Professor Kyoungchul Kong (right) >
Researcher Jeongsu Park, the captain of Team KAIST, said, “We came into this competition thinking of it as a competition against ourselves to begin with and focused on showing the technological gap. Now, we are very happy and proud that our endeavors achieved such good result as well.” He added, “We plan to continue to showcase various functions of the WalkON Suit F1 that have not yet been publicly introduced.”
Researcher Seunghwan Kim, the paraplegic athlete of the team, said, “I am so touched that I was able to introduce the world’s best wearable robot technology of Korea with my own body.”
On a different note, Professor Kong’s research team has successfully commercialized wearable robots through Angel Robotics Co., Ltd. since the 2020 competition. In 2022, they began distributing “ANGEL LEGS M20,” the first wearable robot to be covered by health insurance, and as a result, Angel Robotics Co., Ltd. was successfully listed on KOSDAQ this March.
The various know-how and core technologies accumulated while preparing for this competition is to contribute to further development and propagation of wearable robots, provoking imagination to draw on the future of wearable robots and on how it may change our daily lives.
Final Match (Self-filmed): https://youtu.be/3ASAtvkiOhw
Final Match and Interview (Official Video): https://youtu.be/FSfxOTpDjSE
Final Match and Interview (Summary): https://youtu.be/Sb_vd5-3f_0
2024.10.28 View 1877 -
KAIST Professor Uichin Lee Receives Distinguished Paper Award from ACM
< Photo. Professor Uichin Lee (left) receiving the award >
KAIST (President Kwang Hyung Lee) announced on the 25th of October that Professor Uichin Lee’s research team from the School of Computing received the Distinguished Paper Award at the International Joint Conference on Pervasive and Ubiquitous Computing and International Symposium on Wearable Computing (Ubicomp / ISWC) hosted by the Association for Computing Machinery (ACM) in Melbourne, Australia on October 8.
The ACM Ubiquitous Computing Conference is the most prestigious international conference where leading universities and global companies from around the world present the latest research results on ubiquitous computing and wearable technologies in the field of human-computer interaction (HCI).
The main conference program is composed of invited papers published in the Proceedings of the ACM (PACM) on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT), which covers the latest research in the field of ubiquitous and wearable computing.
The Distinguished Paper Award Selection Committee selected eight papers among 205 papers published in Vol. 7 of the ACM Proceedings (PACM IMWUT) that made outstanding and exemplary contributions to the research community. The committee consists of 16 prominent experts who are current and former members of the journal's editorial board which made the selection after a rigorous review of all papers for a period that stretched over a month.
< Figure 1. BeActive mobile app to promote physical activity to form active lifestyle habits >
The research that won the Distinguished Paper Award was conducted by Dr. Junyoung Park, a graduate of the KAIST Graduate School of Data Science, as the 1st author, and was titled “Understanding Disengagement in Just-in-Time Mobile Health Interventions”
Professor Uichin Lee’s research team explored user engagement of ‘Just-in-Time Mobile Health Interventions’ that actively provide interventions in opportune situations by utilizing sensor data collected from health management apps, based on the premise that these apps are aptly in use to ensure effectiveness.
< Figure 2. Traditional user-requested digital behavior change intervention (DBCI) delivery (Pull) vs. Automatic transmission (Push) for Just-in-Time (JIT) mobile DBCI using smartphone sensing technologies >
The research team conducted a systematic analysis of user disengagement or the decline in user engagement in digital behavior change interventions. They developed the BeActive system, an app that promotes physical activities designed to help forming active lifestyle habits, and systematically analyzed the effects of users’ self-control ability and boredom-proneness on compliance with behavioral interventions over time.
The results of an 8-week field trial revealed that even if just-in-time interventions are provided according to the user’s situation, it is impossible to avoid a decline in participation. However, for users with high self-control and low boredom tendency, the compliance with just-in-time interventions delivered through the app was significantly higher than that of users in other groups.
In particular, users with high boredom proneness easily got tired of the repeated push interventions, and their compliance with the app decreased more quickly than in other groups.
< Figure 3. Just-in-time Mobile Health Intervention: a demonstrative case of the BeActive system: When a user is identified to be sitting for more than 50 mins, an automatic push notification is sent to recommend a short active break to complete for reward points. >
Professor Uichin Lee explained, “As the first study on user engagement in digital therapeutics and wellness services utilizing mobile just-in-time health interventions, this research provides a foundation for exploring ways to empower user engagement.” He further added, “By leveraging large language models (LLMs) and comprehensive context-aware technologies, it will be possible to develop user-centered AI technologies that can significantly boost engagement."
< Figure 4. A conceptual illustration of user engagement in digital health apps. Engagement in digital health apps consists of (1) engagement in using digital health apps and (2) engagement in behavioral interventions provided by digital health apps, i.e., compliance with behavioral interventions. Repeated adherences to behavioral interventions recommended by digital health apps can help achieve the distal health goals. >
This study was conducted with the support of the 2021 Biomedical Technology Development Program and the 2022 Basic Research and Development Program of the National Research Foundation of Korea funded by the Ministry of Science and ICT.
< Figure 5. A conceptual illustration of user disengagement and engagement of digital behavior change intervention (DBCI) apps. In general, user engagement of digital health intervention apps consists of two components: engagement in digital health apps and engagement in behavioral interventions recommended by such apps (known as behavioral compliance or intervention adherence). The distinctive stages of user can be divided into adoption, abandonment, and attrition. >
< Figure 6. Trends of changes in frequency of app usage and adherence to behavioral intervention over 8 weeks, ● SC: Self-Control Ability (High-SC: user group with high self-control, Low-SC: user group with low self-control) ● BD: Boredom-Proneness (High-BD: user group with high boredom-proneness, Low-BD: user group with low boredom-proneness). The app usage frequencies were declined over time, but the adherence rates of those participants with High-SC and Low-BD were significantly higher than other groups. >
2024.10.25 View 1097 -
KAIST Introduces a Wearable Robot that Walks and Puts itself on to Disabled Persons
< Photo 1. WalkON Suit F1 - walking demonstration >
KAIST researchers have unveiled a new wearable robot developed for completely paralyzed persons that can walk to them so that the user can wear it right out of their wheelchairs without the help from others. Also, it was announced that Professor Kyoungchul Kong's team from KAIST will be participating in the wearable robot category of the 3rd Cybathlon, which is being held four years after the team’s gold medal win in 2020.
KAIST (President Kwang-Hyung Lee) announced this new version of the wearable robot for paraplegic people, WalkON Suit F1, by Professor Kyoungchul Kong (CEO and founder of Angel Robotics) of KAIST Department of Mechanical Engineering on the 24th of October.
< Photo 2. (From left) Professor Kyoungchul Kong of the Department of Mechanical Engineering, Researcher Seunghwan Kim (the competing athlete), and PhD candidate Jeongsu Park (the leader of Team KAIST) >
WalkON Suit is a wearable robot for people suffering with paraplegic disabilities that the research team has been continuously researching since 2015. This robot targets to assist persons with American Spinal Injury Association (ASIA) Impairment Scale – A (complete paralysis) grade injury, the most severe level of paraplegia. Therefore, its development purpose is different from that of other rehabilitation therapy and muscle strength assisting robots currently being supplied nationally by Angel Robotics.
Professor Kong's research team first announced WalkON Suit 1 in 2016, and then introduced WalkON Suit 4 in 2020, increasing the walking speed to 3.2 km/h, achieving the normal walking speed of people with no disabilities. In addition, it demonstrated the ability to pass through obstacles such as narrow passages, doors, and stairs that can be encountered in daily life.
However, it had the same fundamental problem all wearable robots have, which is that they require the help of others to wear the robot. While you can walk without help from others once you are wearing the robot, you needed someone's help to put it on to begin with.
The newly released WalkON Suit F1 presented a technical solution to this fundamental problem. It applied a frontal-docking method instead of a rear-sitting method so that you can wear the robot right away without getting out of the wheelchair and into the robot, which would require help from others mid-transition.
< Photo 3. WalkON Suit F1 - suiting-up demonstration >
In addition, before wearing the robot, it can walk on its own like a humanoid robot and approaches the user. It is also implemented a function that actively controls the center of its weight against the pull of gravity so that it maintains balance without falling over even if the user pushes the robot otherwise. The outer design of the WalkON Suit F1, which crosses between a humanoid and a wearable robot, was done by Professor Hyunjoon Park of the Department of Industrial Design at KAIST.
The original function of the wearable robot has also been greatly improved. The performance of the balance control was improved to allow the free use of both hands in upright state, as well as the freedom to take several steps without a cane.
Technological advancements at the components level are also noteworthy. Through close collaboration with Angel Robotics, all core components of the robot, such as the motor, reducer, motor driver, and main circuit, have been domestically produced. The output density of the motor and reducer modules has been improved by about two folds (based on power per weight) compared to the research team's existing technology, and the control performance of the motor driver has been improved by about three times (based on frequency response speed) compared to the best overseas technology.
In particular, the embedded software technology of the motor driver has been significantly improved so that advanced motion control algorithms can be stably implemented without using expensive higher-level controllers. In addition, visual recognition system for obstacle detection and an AI board for neural network application have been installed.
< Figure 1. WalkON Suit F1 shape and main specifications >
Professor Kong explained, “WalkON Suit is the culmination of wearable robot technology for the disabled,” and added, “The numerous components, control, and module technologies derived from WalkON Suit are setting the standard for the entire wearable robot industry.”
Professor Kong’s research team revealed WalkON Suit F1 and announced that they will be participating in the 3rd Cybathlon, which is being held after four years since the last event. In this competition, which will be held on October 27, Professor Kong’s lab, the Exo Lab will be participating with Jeongsu Park, a Ph.D. Student, as the leader and Seunghwan Kim, the lab’s staff researcher with complete paralysis, as the competing athlete.
The difficulty of the missions in this competition has been significantly increased compared to the previous competition, and the number of missions has increased from six to ten. Some missions have been criticized for being overly challenging, going beyond the level that can be encountered in everyday life.
< Photo 4. Cybathlon stadium (Angel Robotics Asia Hub) >
Regarding this, the team leader Jeongsu Park expressed his ambition, saying, “Since we already won first place in the previous competition, our goal in this competition is to show the technological gap rather than competing for rankings.”
The Cybathlon is a cyborg Olympics held every four years in Switzerland. This competition will be held in a hybrid format, with some participants taking part in Switzerland while others broadcasting live from stadiums in their own country on October 27. Professor Kong's research team will be participating via live broadcast from the competition facilities installed in Angel Robotics' Advanced Research Institute (Planet Daejeon).
< Photo 5. Photo of Team KAIST participating in Cybathlon 2024 >
The demonstration video of WalkON Suit F1 can be viewed through the link below.
Link: https://www.youtube.com/@KyoungchulKong_EXO-Lab
2024.10.24 View 3416 -
KAIST Proposes a New Way to Circumvent a Long-time Frustration in Neural Computing
The human brain begins learning through spontaneous random activities even before it receives sensory information from the external world. The technology developed by the KAIST research team enables much faster and more accurate learning when exposed to actual data by pre-learning random information in a brain-mimicking artificial neural network, and is expected to be a breakthrough in the development of brain-based artificial intelligence and neuromorphic computing technology in the future.
KAIST (President Kwang-Hyung Lee) announced on the 23rd of October that Professor Se-Bum Paik 's research team in the Department of Brain Cognitive Sciences solved the weight transport problem*, a long-standing challenge in neural network learning, and through this, explained the principles that enable resource-efficient learning in biological brain neural networks.
*Weight transport problem: This is the biggest obstacle to the development of artificial intelligence that mimics the biological brain. It is the fundamental reason why large-scale memory and computational work are required in the learning of general artificial neural networks, unlike biological brains.
Over the past several decades, the development of artificial intelligence has been based on error backpropagation learning proposed by Geoffery Hinton, who won the Nobel Prize in Physics this year. However, error backpropagation learning was thought to be impossible in biological brains because it requires the unrealistic assumption that individual neurons must know all the connected information across multiple layers in order to calculate the error signal for learning.
< Figure 1. Illustration depicting the method of random noise training and its effects >
This difficult problem, called the weight transport problem, was raised by Francis Crick, who won the Nobel Prize in Physiology or Medicine for the discovery of the structure of DNA, after the error backpropagation learning was proposed by Hinton in 1986. Since then, it has been considered the reason why the operating principles of natural neural networks and artificial neural networks will forever be fundamentally different.
At the borderline of artificial intelligence and neuroscience, researchers including Hinton have continued to attempt to create biologically plausible models that can implement the learning principles of the brain by solving the weight transport problem.
In 2016, a joint research team from Oxford University and DeepMind in the UK first proposed the concept of error backpropagation learning being possible without weight transport, drawing attention from the academic world. However, biologically plausible error backpropagation learning without weight transport was inefficient, with slow learning speeds and low accuracy, making it difficult to apply in reality.
KAIST research team noted that the biological brain begins learning through internal spontaneous random neural activity even before experiencing external sensory experiences. To mimic this, the research team pre-trained a biologically plausible neural network without weight transport with meaningless random information (random noise).
As a result, they showed that the symmetry of the forward and backward neural cell connections of the neural network, which is an essential condition for error backpropagation learning, can be created. In other words, learning without weight transport is possible through random pre-training.
< Figure 2. Illustration depicting the meta-learning effect of random noise training >
The research team revealed that learning random information before learning actual data has the property of meta-learning, which is ‘learning how to learn.’ It was shown that neural networks that pre-learned random noise perform much faster and more accurate learning when exposed to actual data, and can achieve high learning efficiency without weight transport.
< Figure 3. Illustration depicting research on understanding the brain's operating principles through artificial neural networks >
Professor Se-Bum Paik said, “It breaks the conventional understanding of existing machine learning that only data learning is important, and provides a new perspective that focuses on the neuroscience principles of creating appropriate conditions before learning,” and added, “It is significant in that it solves important problems in artificial neural network learning through clues from developmental neuroscience, and at the same time provides insight into the brain’s learning principles through artificial neural network models.”
This study, in which Jeonghwan Cheon, a Master’s candidate of KAIST Department of Brain and Cognitive Sciences participated as the first author and Professor Sang Wan Lee of the same department as a co-author, will be presented at the 38th Neural Information Processing Systems (NeurIPS), the world's top artificial intelligence conference, to be held in Vancouver, Canada from December 10 to 15, 2024. (Paper title: Pretraining with random noise for fast and robust learning without weight transport)
This study was conducted with the support of the National Research Foundation of Korea's Basic Research Program in Science and Engineering, the Information and Communications Technology Planning and Evaluation Institute's Talent Development Program, and the KAIST Singularity Professor Program.
2024.10.23 View 980 -
KAIST Develops Technology for the Precise Diagnosis of Electric Vehicle Batteries Using Small Currents
Accurately diagnosing the state of electric vehicle (EV) batteries is essential for their efficient management and safe use. KAIST researchers have developed a new technology that can diagnose and monitor the state of batteries with high precision using only small amounts of current, which is expected to maximize the batteries’ long-term stability and efficiency.
KAIST (represented by President Kwang Hyung Lee) announced on the 17th of October that a research team led by Professors Kyeongha Kwon and Sang-Gug Lee from the School of Electrical Engineering had developed electrochemical impedance spectroscopy (EIS) technology that can be used to improve the stability and performance of high-capacity batteries in electric vehicles.
EIS is a powerful tool that measures the impedance* magnitude and changes in a battery, allowing the evaluation of battery efficiency and loss. It is considered an important tool for assessing the state of charge (SOC) and state of health (SOH) of batteries. Additionally, it can be used to identify thermal characteristics, chemical/physical changes, predict battery life, and determine the causes of failures. *Battery Impedance: A measure of the resistance to current flow within the battery that is used to assess battery performance and condition.
However, traditional EIS equipment is expensive and complex, making it difficult to install, operate, and maintain. Moreover, due to sensitivity and precision limitations, applying current disturbances of several amperes (A) to a battery can cause significant electrical stress, increasing the risk of battery failure or fire and making it difficult to use in practice.
< Figure 1. Flow chart for diagnosis and prevention of unexpected combustion via the use of the electrochemical impedance spectroscopy (EIS) for the batteries for electric vehicles. >
To address this, the KAIST research team developed and validated a low-current EIS system for diagnosing the condition and health of high-capacity EV batteries. This EIS system can precisely measure battery impedance with low current disturbances (10mA), minimizing thermal effects and safety issues during the measurement process.
In addition, the system minimizes bulky and costly components, making it easy to integrate into vehicles. The system was proven effective in identifying the electrochemical properties of batteries under various operating conditions, including different temperatures and SOC levels.
Professor Kyeongha Kwon (the corresponding author) explained, “This system can be easily integrated into the battery management system (BMS) of electric vehicles and has demonstrated high measurement accuracy while significantly reducing the cost and complexity compared to traditional high-current EIS methods. It can contribute to battery diagnosis and performance improvements not only for electric vehicles but also for energy storage systems (ESS).”
This research, in which Young-Nam Lee, a doctoral student in the School of Electrical Engineering at KAIST participated as the first author, was published in the prestigious international journal IEEE Transactions on Industrial Electronics (top 2% in the field; IF 7.5) on September 5th. (Paper Title: Small-Perturbation Electrochemical Impedance Spectroscopy System With High Accuracy for High-Capacity Batteries in Electric Vehicles, Link: https://ieeexplore.ieee.org/document/10666864)
< Figure 2. Impedance measurement results of large-capacity batteries for electric vehicles. ZEW (commercial EW; MP10, Wonatech) versus ZMEAS (proposed system) >
This research was supported by the Basic Research Program of the National Research Foundation of Korea, the Next-Generation Intelligent Semiconductor Technology Development Program of the Korea Evaluation Institute of Industrial Technology, and the AI Semiconductor Graduate Program of the Institute of Information & Communications Technology Planning & Evaluation.
2024.10.17 View 1345 -
KAIST ISSS Research Session Captivates 150↑ International Scholars, Achieve Major Success
< Photo. Scholars gatheres for NRF Information Session at Chung Keun Mo Hall >
KAIST’s International Office, headed by Vice President Soyoung Kim, successfully organized the ‘NRF Information Session for International Scholars’ on September 11, 2024, in collaboration with the National Research Foundation of Korea (NRF).
The event was held at KAIST’s main campus to enourage the international scholar’s active participation in research projects and support their establishment of stable research environment and integration into Korea’s academic community by introducing NRF’s key research programs.
Divided into two main segments – science and engineering, and humanities and social sciences – the session attracted approximately 150 international faculty and researchers from 23 universities across the nation.
The event commenced with a keynote address by Vice President Soyoung Kim, followed by a presentation from Dr. Seol Min of the National Research Foundation, who highlighted basic research initiatives in the science and technology sector. Subsequently, Professor Daniel Martin from the Digital Humanities and Social Sciences Department and Professor Thomas Steinberger from the Department of Business and Technology Management presented practical research project support case studies, sharing invaluable insights gained from their domestic research experiences.
Following the information session, participants engaged in a networking event, where researchers involved in major R&D projects exchanged insights and discussed their ongoing research initiatives.
An international professor remarked, “My understanding of NRF’s research programs for international researchers has broadened considerably. I am now more inclined to actively participate in projects organized by NRF in the future.”
Vice President Kim expressed her aspiration that the event would address the challenges faced by researchers and offer essential support to those engaged in research projects. “We will stay attuned to the needs of the research community and work towards creating a more supportive research environment,” said the VP.
Meanwhile, KAIST hosts a distinguished faculty comprising 134 professors from 22 countries and 71 researchers representing 23 nations, all contributing to groundbreaking academic achievements. Additionally, KAIST is home to over 1,000 international students from more than 100 countries, actively pursuing their studies. This diverse composition of global talent reinforces KAIST's position as a leading international hub for research and education.
2024.09.13 View 1235 -
KAIST and NYU set out to Install Korea's First Joint Degree Program in AI
< (From left) New York University President Linda Mills and President Kwang-Hyung Lee >
KAIST (President Kwang-Hyung Lee) and New York University (NYU, President Linda G. Mills) signed an MOU in the afternoon of the 9th to introduce a graduate program for a joint degree in the field of artificial intelligence.
This agreement was promoted based on the consensus between the two universities that strengthening capabilities in the field of AI and fostering global talent are essential elements that can lead to great development in the entire future society beyond simple technical education.
The two universities have been operating joint research groups in various industrial fields related to AI and convergence with it, and based on this agreement, they plan to establish an operating committee within this year to design a joint degree program for graduate school courses related to artificial intelligence.
A KAIST official said, “If the joint degree program in AI is implemented, it is expected to be an unprecedented innovative experiment in which KAIST and NYU join forces to create ‘a single AI degree.’
The committee will consist of an equal number of faculty members from both schools, and will discuss the overall strategic planning of the joint degree program, including ▴curriculum structure and course composition ▴course completion roadmap ▴calculation of faculty and student population ▴calculation of budget size ▴calculation of operating facility size and details ▴legal matters regarding certification. In addition, the development of a new logo symbolizing the joint degree of KAIST and NYU in AI will also be carried out.
The two schools expect that the joint degree program being promoted this time will contribute to advancing education and research capabilities in the field of artificial intelligence, jointly discovering and fostering talent in related fields that are currently lacking worldwide, and will become an exemplary case of global education and research cooperation.
The faculty members of both schools, who possess excellent capabilities, will provide innovative and creative education in the field of artificial intelligence. Students will receive support to gain top-level research experience by participating in various international joint research projects promoted by the faculty members of both schools. Through this, the core of this joint degree program promoted by both schools is to continuously cultivate excellent human resources who will lead the future global society.
Since signing a cooperation agreement for the establishment of a joint campus in June 2022, KAIST and NYU have been promoting campus sharing, joint research, and joint bachelor's degree programs. Including this, they are developing an innovative joint campus model and establishing an active international cooperation model.
In particular, the exchange student system for undergraduate students will be implemented starting from the second semester of the 2023 academic year. 30 students from KAIST and 11 students from NYU were selected through a competitive selection process and are participating. In the case of KAIST students, if they complete one of the six minor programs at NYU, they will receive a degree that states the completion of the minor upon graduation.
Based on the performance of the undergraduate exchange student operation, the two schools have also agreed to introduce a dual degree system for master's and doctoral students, and specific procedures are currently in progress.
In addition, from 2023 to the present, we are carrying out future joint research projects in 15 fields that are integrated with AI, and we plan to begin international joint research in 10 fields centered on AI and bio from the fourth quarter of this year.
NYU President Linda Mills said, “AI technology can play a significant role in addressing various social challenges such as climate change, health care, and education inequality,” and added that, “The global talent cultivated through our two schools will also go on to make innovative contributions to solving these social problems.”
Kwang-Hyung Lee, the president of KAIST, said, “In the era of competition for global hegemony in technology, the development of AI technology is an essential element for countries and companies to secure competitiveness,” and “Through long-term cooperation with NYU, we will take the lead in fostering world-class, advanced talents who can innovatively apply and develop AI in various fields.”
The signing ceremony held at the Four Seasons Hotel in Seoul was attended by KAIST officials including President Kwang-Hyung Lee, Hyun Deok Yeo, the Director of G-School, NYU officials including President Linda Mills, Kyunghyun Cho, a Professor of Computer Science and Data Science, and Dr. Karin Pavese, the Executive Director of NYU-KAIST Innovation Research Institute, amid attendance by other key figures from the industries situated in Korea. (End)
2024.09.10 View 1909 -
KAIST presents strategies for Holotomography in advanced bio research
Measuring and analyzing three-dimensional (3D) images of live cells and tissues is considered crucial in advanced fields of biology and medicine. Organoids, which are 3D structures that mimic organs, are particular examples that significantly benefits 3D live imaging. Organoids provide effective alternatives to animal testing in the drug development processes, and can rapidly determine personalized medicine. On the other hand, active researches are ongoing to utilize organoids for organ replacement.
< Figure 1. Schematic illustration of holotomography compared to X-ray CT. Similar to CT, they share the commonality of measuring the optical properties of an unlabeled specimen in three dimensions. Instead of X-rays, holotomography irradiates light in the visible range, and provides refractive index measurements of transparent specimens rather than absorptivity. While CT obtains three-dimensional information only through mechanical rotation of the irradiating light, holotomography can replace this by applying wavefront control technology in the visible range. >
Organelle-level observation of 3D biological specimens such as organoids and stem cell colonies without staining or preprocessing holds significant implications for both innovating basic research and bioindustrial applications related to regenerative medicine and bioindustrial applications.
Holotomography (HT) is a 3D optical microscopy that implements 3D reconstruction analogous to that of X-ray computed tomography (CT). Although HT and CT share a similar theoretical background, HT facilitates high-resolution examination inside cells and tissues, instead of the human body. HT obtains 3D images of cells and tissues at the organelle level without chemical or genetic labeling, thus overcomes various challenges of existing methods in bio research and industry. Its potential is highlighted in research fields where sample physiology must not be disrupted, such as regenerative medicine, personalized medicine, and infertility treatment.
< Figure 2. Label-free 3D imaging of diverse live cells. Time-lapse image of Hep3B cells illustrating subcellular morphology changes upon H2O2 treatment, followed by cellular recovery after returning to the regular cell culture medium. >
This paper introduces the advantages and broad applicability of HT to biomedical researchers, while presenting an overview of principles and future technical challenges to optical researchers. It showcases various cases of applying HT in studies such as 3D biology, regenerative medicine, and cancer research, as well as suggesting future optical development. Also, it categorizes HT based on the light source, to describe the principles, limitations, and improvements of each category in detail. Particularly, the paper addresses strategies for deepening cell and organoid studies by introducing artificial intelligence (AI) to HT.
Due to its potential to drive advanced bioindustry, HT is attracting interest and investment from universities and corporates worldwide. The KAIST research team has been leading this international field by developing core technologies and carrying out key application researches throughout the last decade.
< Figure 3. Various types of cells and organelles that make up the imaging barrier of a living intestinal organoid can be observed using holotomography. >
This paper, co-authored by Dr. Geon Kim from KAIST Research Center for Natural Sciences, Professor Ki-Jun Yoon's team from the Department of Biological Sciences, Director Bon-Kyoung Koo's team from the Institute for Basic Science (IBS) Center for Genome Engineering, and Dr. Seongsoo Lee's team from the Korea Basic Science Institute (KBSI), was published in 'Nature Reviews Methods Primers' on the 25th of July. This research was supported by the Leader Grant and Basic Science Research Program of the National Research Foundation, the Hologram Core Technology Development Grant of the Ministry of Science and ICT, the Nano and Material Technology Development Project, and the Health and Medical R&D Project of the Ministry of Health and Welfare.
2024.07.30 View 1586 -
Unraveling Mitochondrial DNA Mutations in Human Cells
Throughout our lifetime, cells accumulate DNA mutations, which contribute to genetic diversity, or “mosaicism”, among cells. These genomic mutations are pivotal for the aging process and the onset of various diseases, including cancer. Mitochondria, essential cellular organelles involved in energy metabolism and apoptosis, possess their own DNA, which are susceptible to mutations. However, studies on mtDNA mutations and mosaicism have been limited due to a variety of technical challenges.
Genomic scientists from KAIST have revealed the genetic mosaicism characterized by variations in mitochondrial DNA (mtDNA) across normal human cells. This study provides fundamental insights into understanding human aging and disease onset mechanisms.
The study, “Mitochondrial DNA mosaicism in normal human somatic cells,” was published in Nature Genetics on July 22. It was led by graduate student Jisong An under the supervision of Professor Young Seok Ju from the Graduate School of Medical Science and Engineering.
Researchers from Seoul National University College of Medicine, Yonsei University College of Medicine, Korea University College of Medicine, Washington University School of Medicine National Cancer Center, Seoul National University Hospital, Gangnam Severance Hospital and KAIST faculty startup company Inocras Inc. also participated in this study.
< Figure 1. a. Flow of experiment. b. Schematic diagram illustrating the origin and dynamics of mtDNA alterations across a lifetime. >
The study involved a bioinformatic analysis of whole-genome sequences from 2,096 single cells obtained from normal human colorectal epithelial tissue, fibroblasts, and blood collected from 31 individuals. The study highlights an average of three significant mtDNA differences between cells, with approximately 6% of these variations confirmed to be inherited as heteroplasmy from the mother.
Moreover, mutations significantly increased during tumorigenesis, with some mutations contributing to instability in mitochondrial RNA. Based on these findings, the study illustrates a computational model that comprehensively elucidates the evolution of mitochondria from embryonic development to aging and tumorigenesis.
This study systematically reveals the mechanisms behind mitochondrial DNA mosaicism in normal human cells, establishing a crucial foundation for understanding the impact of mtDNA on aging and disease onset.
Professor Ju remarked, “By systematically utilizing whole-genome big data, we can illuminate previously unknown phenomena in life sciences.” He emphasized the significance of the study, adding, “For the first time, we have established a method to systematically understand mitochondrial DNA changes occurring during human embryonic development, aging, and cancer development.”
This work was supported by the National Research Foundation of Korea and the Suh Kyungbae Foundation.
2024.07.24 View 1927 -
The 3rd Global Entrepreneurship Summer School (GESS 2024) Successfully Completed in Silicon Valley
The 2024 Global Entrepreneurship Summer School (2024 KAIST GESS), hosted by the Office of Global Initiatives under the KAIST International Office (Director Man-Sung Yim), was held for the third time. This program allows students to visit Silicon Valley, a global startup hub, to directly experience its famous startup ecosystem and develop their capabilities for global expansion. A total of 20 students were selected through applications, interviews, final presentations, mentoring, and peer evaluations. Additionally, 17 students from the KAIST Impact MBA course at the KAIST Business School also participated.
Before starting the Silicon Valley program, participants received mentoring on business model development and pitching advice from a senior entrepreneur at KAIST for about two months, beginning last May. Afterward, they developed business items for each team at KAIST’s main campus in Daejeon. For seven days, starting from June 23rd, workshops were held under the themes of global entrepreneurship, learning through failure, capital and network, and startup culture at KOTRA Silicon Valley Trade Center, JP Morgan, and Plug and Play Tech Center. This program's lecture series provided prospective entrepreneurs with the opportunity to systematically learn the mindset and gain the experience needed to start a global business.
The participants also visited local companies and gained experience in the field of global technology startups. Visits included Bear Robotics (CEO John Ha), Soundable Health (CEO Cathering Song), ImpriMed (CEO Sungwon Lim), Phantom AI (CEO Hyunggi Cho), B Garage (CEO Aiden Kim), and Simple Steps (CEO Doyeon Kim). Lectures contained vivid experiences from Silicon Valley CEOs and company tours boosted the students' passion for entrepreneurship. In particular, Doyeon Kim, CEO of Simple Steps, which helps prevent career breaks for Korean female immigrants in Silicon Valley and allows talented female immigrants to demonstrate their abilities in society, said, “As a KAIST alumna entrepreneur, it was meaningful to share my experience with this generation of students who dream of starting a global business and creating social enterprises in the United States.”
This program also included a tour of Silicon Valley's big tech companies that have made a significant impact on the digital ecosystem through technological advancement and innovation. This included Broadcom, which maintains a strong global presence in the semiconductor and infrastructure software technology fields. At the invitation of Chairman Hock Tan, GESS participants had the opportunity to attend his lecture and ask questions. Chairman Tan, who received an honorary doctorate in engineering from KAIST last February, emphasized that experiencing failure and giving consistent effort over a long period of time are more important than anything else in order to grow as a global entrepreneur, and that technologies influencing the global market evolve over generations.
< Photo. Group photo of GESS 2024 participants at Broadcom with Chairman Hock Tan (center) ⓒBroadcom>
As part of this program, participants conducted a volunteer program called 'Let's play with AI+ Tech' with the Sunnyvale community in Silicon Valley and Foothill College to help grow together with the community. Through this program, GESS participants cultivated the virtues of a global leader. In this volunteer activity, low-income elementary school students and parents from the Sunnyvale community participated in chatbot training led by KAIST students, providing an opportunity to work with underprivileged groups in the local community.
In the final pitching event, the highlight of the program, local venture investors from Silicon Valley were invited as judges and evaluated the pitches for each team's business items. The participating students, who developed their own business models while receiving advice through face-to-face mentoring from a professional accelerator in Silicon Valley, showcased their creative and innovative ideas, presenting themselves as future global entrepreneurs. Merey Makhmutova (BS in Civil and Environmental Engineering) from the K-Bridge team, who won the final pitch, expressed her ambition: “Even before GESS pitch day, our team kept refining the pitch deck as we attended the lectures and benefitted from the mentoring. Our intense teamwork was a significant reason why we ultimately won first prize.” She added that K-Bridge aims to win an award at the upcoming UKC Pitching Competition and expressed her gratitude for being able to participate in this program. Arseniy Kan (BS in Electrical Engineering) from the KAIST Enablers team, who took second place, said, “The 2024 KAIST GESS Program became the most unforgettable and precious opportunity of my lifetime, and I dream of using this opportunity as a stepping stone to becoming a global entrepreneur.“ Additionally, Kangster (CEO Kang Kim), who won the Impact MBA final pitching session, had the opportunity to secure a meeting with a local investment company after their GESS final pitch.
The 2024 KAIST GESS was held in cooperation with the KAIST International Office, the KAIST College of Business, and Startup KAIST. Director Man-Sung Yim from the Office of Global Initiatives, who hosted the event, said, “KAIST students will grow into leaders with global influence and contribute to the international community by creating global value. At the same time, we hope to raise the international status of our university.” Professor Sangchan Park, who led the 17 Impact MBA students in this educational program, added, “Meeting with companies leading the global market and visiting Silicon Valley has been a valuable learning experience for students aiming to start a global startup.”
KAIST plans to continue promoting its global entrepreneurship education program by enriching its curriculum each year and helping students grow into entrepreneurs with the virtues of global leaders.
2024.07.03 View 3225