Artificial+Intelligence
-
‘Carrier-Resolved Photo-Hall’ to Push Semiconductor Advances
(Professor Shin and Dr. Gunawan (left))
An IBM-KAIST research team described a breakthrough in a 140-year-old mystery in physics. The research reported in Nature last month unlocks the physical characteristics of semiconductors in much greater detail and aids in the development of new and improved semiconductor materials.
Research team under Professor Byungha Shin at the Department of Material Sciences and Engineering and Dr. Oki Gunawan at IBM discovered a new formula and technique that enables the simultaneous extraction of both majority and minority carrier information such as their density and mobility, as well as gain additional insights about carrier lifetimes, diffusion lengths, and the recombination process. This new discovery and technology will help push semiconductor advances in both existing and emerging technologies.
Semiconductors are the basic building blocks of today’s digital electronics age, providing us with a multitude of devices that benefit our modern life. To truly appreciate the physics of semiconductors, it is very important to understand the fundamental properties of the charge carriers inside the materials, whether those particles are positive or negative, their speed under an applied electric field, and how densely they are packed into the material.
Physicist Edwin Hall found a way to determine those properties in 1879, when he discovered that a magnetic field will deflect the movement of electronic charges inside a conductor and that the amount of deflection can be measured as a voltage perpendicular to the flow of the charge. Decades after Hall’s discovery, researchers also recognized that they can measure the Hall effect with light via “photo-Hall experiments”. During such experiments, the light generates multiple carriers or electron–hole pairs in the semiconductors.
Unfortunately, the basic Hall effect only provided insights into the dominant charge carrier (or majority carrier). Researchers were unable to extract the properties of both carriers (the majority and minority carriers) simultaneously. The property information of both carriers is crucial for many applications that involve light such as solar cells and other optoelectronic devices.
In the photo-Hall experiment by the KAIST-IBM team, both carriers contribute to changes in conductivity and the Hall coefficient. The key insight comes from measuring the conductivity and Hall coefficient as a function of light intensity. Hidden in the trajectory of the conductivity, the Hall coefficient curve reveals crucial new information: the difference in the mobility of both carriers. As discussed in the paper, this relationship can be expressed elegantly as: Δµ = d (σ²H)/dσ
The research team solved for both majority and minority carrier mobility and density as a function of light intensity, naming the new technique Carrier-Resolved Photo Hall (CRPH) measurement. With known light illumination intensity, the carrier lifetime can be established in a similar way.
Beyond advances in theoretical understanding, advances in experimental techniques were also critical for enabling this breakthrough. The technique requires a clean Hall signal measurement, which can be challenging for materials where the Hall signal is weak due to low mobility or when extra unwanted signals are present, such as under strong light illumination.
The newly developed photo-Hall technique allows the extraction of an astonishing amount of information from semiconductors. In contrast to only three parameters obtained in the classic Hall measurements, this new technique yields up to seven parameters at every tested level of light intensity. These include the mobility of both the electron and hole; their carrier density under light; the recombination lifetime; and the diffusion lengths for electrons, holes, and ambipolar types. All of these can be repeated N times (i.e. the number of light intensity settings used in the experiment).
Professor Shin said, “This novel technology sheds new light on understanding the physical characteristics of semiconductor materials in great detail.” Dr. Gunawan added, “This will will help accelerate the development of next-generation semiconductor technology such as better solar cells, better optoelectronics devices, and new materials and devices for artificial intelligence technology.”
Profile:
Professor Byungha Shin
Department of Materials Science and Engineering
KAIST
byungha@kaist.ac.kr
http://energymatlab.kaist.ac.kr/
2019.11.18 View 15742 -
AI to Determine When to Intervene with Your Driving
(Professor Uichin Lee (left) and PhD candidate Auk Kim)
Can your AI agent judge when to talk to you while you are driving? According to a KAIST research team, their in-vehicle conservation service technology will judge when it is appropriate to contact you to ensure your safety.
Professor Uichin Lee from the Department of Industrial and Systems Engineering at KAIST and his research team have developed AI technology that automatically detects safe moments for AI agents to provide conversation services to drivers.
Their research focuses on solving the potential problems of distraction created by in-vehicle conversation services. If an AI agent talks to a driver at an inopportune moment, such as while making a turn, a car accident will be more likely to occur.
In-vehicle conversation services need to be convenient as well as safe. However, the cognitive burden of multitasking negatively influences the quality of the service. Users tend to be more distracted during certain traffic conditions. To address this long-standing challenge of the in-vehicle conversation services, the team introduced a composite cognitive model that considers both safe driving and auditory-verbal service performance and used a machine-learning model for all collected data.
The combination of these individual measures is able to determine the appropriate moments for conversation and most appropriate types of conversational services. For instance, in the case of delivering simple-context information, such as a weather forecast, driver safety alone would be the most appropriate consideration. Meanwhile, when delivering information that requires a driver response, such as a “Yes” or “No,” the combination of driver safety and auditory-verbal performance should be considered.
The research team developed a prototype of an in-vehicle conversation service based on a navigation app that can be used in real driving environments. The app was also connected to the vehicle to collect in-vehicle OBD-II/CAN data, such as the steering wheel angle and brake pedal position, and mobility and environmental data such as the distance between successive cars and traffic flow.
Using pseudo-conversation services, the research team collected a real-world driving dataset consisting of 1,388 interactions and sensor data from 29 drivers who interacted with AI conversational agents. Machine learning analysis based on the dataset demonstrated that the opportune moments for driver interruption could be correctly inferred with 87% accuracy.
The safety enhancement technology developed by the team is expected to minimize driver distractions caused by in-vehicle conversation services. This technology can be directly applied to current in-vehicle systems that provide conversation services. It can also be extended and applied to the real-time detection of driver distraction problems caused by the use of a smartphone while driving.
Professor Lee said, “In the near future, cars will proactively deliver various in-vehicle conversation services. This technology will certainly help vehicles interact with their drivers safely as it can fairly accurately determine when to provide conversation services using only basic sensor data generated by cars.”
The researchers presented their findings at the ACM International Joint Conference on Pervasive and Ubiquitous Computing (Ubicomp’19) in London, UK. This research was supported in part by Hyundai NGV and by the Next-Generation Information Computing Development Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT.
(Figure: Visual description of safe enhancement technology for in-vehicle conversation services)
2019.11.13 View 19664 -
KAIST to Transfer Core Tech to Domestic Companies amid Japan's Export Curbs
< Associate Vice President Kyung-Cheol Choi of the Office of University-Industry Cooperation (OUIC) at KAIST >
KAIST will transfer four core technologies related to materials, parts, and equipment to domestic companies to help them combat the latest export curbs triggered by Korea’s removal from Japan’s ‘white list’ of preferential trade partners. In addition, KAIST’s five patented technologies in the field of artificial intelligence (AI) and materials and parts will also be transferred to the companies in order to reduce the reliance on Japan and achieve technological independence through the ‘localization’ of key technologies.
KAIST announced these university-industry cooperation promotion plans at the ‘2019 KAIST Core Tech Transfer Day Conference’ held in Seoul on September 17. More than 200 entrepreneurs and investors attended the briefing and on-site consulting sessions delivered by nine KAIST professors who led the development of the technologies.
The four technologies were presented at the conference as those that can replace Japanese technologies subject to the export curbs. They include:
1. ‘Transparent fluorinated polyimide with low thermal expansion’ developed by Professor Sang-Youl Kim of the Department of Chemistry
2. ‘A non-destructive electromagnetic performance testing system’ developed by Professor Jung-Ryul Lee of the Department of Aerospace Engineering
3. ‘A nanotechnology-based electrode material for use in advanced secondary batteries’ developed by Professor Do-Kyung Kim of the Department of Materials Science and Engineering
4. ‘A high-resolution photoresist’ developed by Professor Emeritus Jin-Baek Kim of the Department of Chemistry.
Of particular interest is the non-destructive electromagnetic performance testing system technology developed by Professor Jung-Ryul Lee. This new cost-effective technology enables tests that were impossible to carry out using conventional technologies and yields a cost reduction of more than 50 percent compared to foreign technologies.
By introducing Professor Do-Kyung Kim’s new electrode material technology, the efficiency of electric vehicles can be increased. As this technology uses relatively low-cost sodium ion batteries, industries can prepare for the possible jump from the more expensive lithium batteries currently being used.
Another five patented AI and materials and parts technologies disclosed at the conference include:
1. ‘Enhanced HTTP adaptive streaming with CNN-based super-resolution’ developed by Professor Dong-soo Han of the School of Electrical Engineering
2. ‘Method and apparatus of brain-computer interface design for estimating choice behavior and decision strategy’ developed by Professor Sang-Wan Lee of the Department of Bio and Brain Engineering
3. ‘Eco-friendly fabrication of metal oxide nanoparticles and fabrication of non-toxic polymer sunscreen ingredients by electron irradiation’ developed by Professor Sung-Oh Cho of the Department of Nuclear and Quantum Engineering
4. ‘High-density nanofiber yarn-based coloricmetric gas sensors’ developed by Professor Il-Doo Kim of the Department of Materials Science and Engineering
5. ‘Silicon-pocket energy storage electrode with high energy density and its manufacturing technology’ developed by Professor Jeung-Ku Kang of the Graduate school of EEWS.
The patented nanofiber-based coloricmetric gas sensor technology developed by Professor Il-Doo Kim allows for the diagnosis of diseases by only using the patient’s respiration. Due to its high productivity and processability, it is expected to be applied to various fields in the fast-growing disease diagnosis sensor market, which includes mobile devices and wearable sensors.
Moreover, Professor Dong-soo Han’s patented adaptive streaming technology attracted attention along with the ever-growing Over The Top (OTT) and Video On Demand (VOD) service markets, since it has significant potential for improving the streaming quality of videos and reducing costs for video providers.
Professor Kyung-Cheol Choi, the Associate Vice President of the Office of University-Industry Cooperation (OUIC) at KAIST, said, “KAIST OUIC and KAIST Advisors on Materials and Parts (KAMP) have been working tirelessly to help Korean companies cope with the recent Japanese export restrictions. KAIST’s efforts will enhance the competitiveness and growth of the Korean industry and economy, turning this national crisis into opportunity.”
(END)
2019.09.20 View 7898 -
FIRIC-EU JRC Joint Workshop on Smart Specialization
The Fourth Industrial Revolution Intelligence Center (FIRIC) at KAIST discussed ‘Smart Specialization’ for regional innovation and economic growth in the wake of the Fourth Industrial Revolution during the workshop with the EU Joint Research Center (EU-JRC) in Seville, Spain last week. The two sides also agreed to sign an MOU to expand mutual collaboration.
KAIST’s FIRIC was founded in cooperation with the World Economic Forum in July 2017 to carry out policy research for the promotion of science and technology-based inclusive growth and innovation and to lead related global efforts. The EU-JRC has committed to developing cohesive policies that aim to narrow regional gaps within the European Union. Founded in 1958 in Brussels, the EU-JRC has long been in charge of EU strategies for regional innovation based on emerging technologies.
The workshop also covered issues related to public-private partnerships and innovation clusters from the perspective of the EU and Asia, such as the global value chain and the implementation of industrial clusters policy amid the changes in the industrial ecosystem due to digitalization, automation, and the utilization of robotics during the Fourth Industrial Revolution.
In addition, the session included discussions on inclusive growth and job market changes in the era of the Fourth Industrial Revolution, addressing how Smart Specialization and the outcomes of the 4IR will shift the paradigm of current job and technology capabilities, as well as employment issues in many relevant industries. In particular, the actual case studies and their related policies and regulatory trends regarding the potential risks and ethical issues of artificial intelligence were introduced.
Regarding the financial services that utilize blockchain technologies and the establishment of public sector governance for such technologies, the participating experts noted difficulties in the diffusion of blockchain-based local currencies or public services, which call for a sophisticated analytical and practical framework for innovative and transparent governance.
Dr. Mark Boden, the Team Leader of the EU-JRC, introduced the EU’s initiatives to promote Smart Specialization, such as its policy process, governance design, vision sharing, and priority setting, with particular emphasis on targeted support for Smart Specialization in lagging regions. Professor So Young Kim, who is the dean of the Graduate School of Science and Technology Policy and FIRIC’s Deputy Director said, “KAIST’s global role regarding the Fourth Industrial Revolution will be expanded in the process of exploring and developing innovative models of technology-policy governance while working jointly with the EU-JRC.”
2019.08.02 View 8411 -
Education Innovation Day Reaffirms Rewarding of Excellence
Professors Tae-Eog Lee and Il-Chul Moon from the Department of Industrial & Systems Engineering received the Linkgenesis Best Teacher Award and the Soo-Young Lee Teaching Innovation Award on May 10. They were each awarded with 10 million KRW in prize money during the Education Innovation Day ceremony held at the Chung Kun-mo conference hall.
The award was endowed by KAIST Alumni Scholarship Chairman Hyung-Kyu Lim and KAIST Foundation Chairman Soo-Young Lee to support the innovation initiative and acknowledge faculty members who made significant contributions to educational innovation and benefited the general public though their innovations.
“KAIST’s vision for excellence and commitment to innovation is a game changer. Educational innovation is one of five pillars of Vision 2031, and it is our priority to foster critical and creative thinking students,” said President Sung-Chul Shin at the ceremony. All the awardees made presentation on their innovative projects and shared their ideas on better pedagogical methodology for next generation.
Professor Lee, dean of the KAIST Academy and the head of the Center for Excellence in Learning & Teaching was recognized for his contribution to enhancing educational quality through innovative learning and teaching methodology development. He has set up an Education 3.0 Initiative, an online education platform for flipped learning at KAIST.
Professor Moon also upgraded the online education platform to the 4.0 version and extended KAIST’s massive online courses through KOOC framework. This open platform offers more than 62 courses, with more than 170 thousand users registered since 2014.
Professor Song-Hong Park from the Department of Bio and Brain Engineering and Professor Jae-Woo Lee from the Department of Chemical and Biomolecular Engineering also won the Excellence Award.
2019.05.10 View 10151 -
Professor Ji-Hyun Lee Awarded the Sasada Prize
Professor Ji-Hyun Lee from the Graduate School of Culture Technology was awarded the Sasada Prize during the 24th annual Conference of Computer-Aided Architectural Design Research in Asia (CAADRIA) held in Wellington, New Zealand on April 15.
The Sasada Award honors the late Professor Tsuyoshi Sasada (1941-2005), the former Professor of Osaka University and co-founder and fellow of CAADRIA. It is given to an individual who has contributed to the next generation of researchers and academics, to the wider profession and practice in computer-aided design and research, and has earned recognition in the academic community.
Professor Lee was recognized for her development of CAAD (Computer-Aided Architectural Design) through her research work on the land price precision system using case-based reasoning. Her research team proposed a model for estimating the average apartment price in an administrative district after collecting 40 variables from the six major Korean cities, excluding Seoul and Ulsan. Their follow-up studies showed the possibility of replacing existing experts’ predictions.
Professor Lee has been steadily researching for 20 years on case-based reasoning (CBR), a field of artificial intelligence, and has published more than 40 papers in the field of CBR. Meanwhile, the CAAD Future 2019 event will be held at KAIST in June.
2019.04.23 View 8159 -
Brain-inspired Artificial Intelligence in Robots
(from left: PhD candidate Su Jin An, Dr. Jee Hang Lee and Professor Sang Wan Lee)
Research groups in KAIST, the University of Cambridge, Japan’s National Institute for Information and Communications Technology, and Google DeepMind argue that our understanding of how humans make intelligent decisions has now reached a critical point in which robot intelligence can be significantly enhanced by mimicking strategies that the human brain uses when we make decisions in our everyday lives.
In our rapidly changing world, both humans and autonomous robots constantly need to learn and adapt to new environments. But the difference is that humans are capable of making decisions according to the unique situations, whereas robots still rely on predetermined data to make decisions.
Despite the rapid progress being made in strengthening the physical capability of robots, their central control systems, which govern how robots decide what to do at any one time, are still inferior to those of humans. In particular, they often rely on pre-programmed instructions to direct their behavior, and lack the hallmark of human behavior, that is, the flexibility and capacity to quickly learn and adapt.
Applying neuroscience in robotics, Professor Sang Wan Lee from the Department of Bio and Brain Engineering, KAIST and Professor Ben Seymour from the University of Cambridge and Japan’s National Institute for Information and Communications Technology proposed a case in which robots should be designed based on the principles of the human brain. They argue that robot intelligence can be significantly enhanced by mimicking strategies that the human brain uses during decision-making processes in everyday life.
The problem with importing human-like intelligence into robots has always been a difficult task without knowing the computational principles for how the human brain makes decisions –in other words, how to translate brain activity into computer code for the robots’ ‘brains’.
However, researchers now argue that, following a series of recent discoveries in the field of computational neuroscience, there is enough of this code to effectively write it into robots. One of the examples discovered is the human brain’s ‘meta-controller’, a mechanism by which the brain decides how to switch between different subsystems to carry out complex tasks. Another example is the human pain system, which allows them to protect themselves in potentially hazardous environments. “Copying the brain’s code for these could greatly enhance the flexibility, efficiency, and safety of robots,” Professor Lee said.
The team argued that this inter-disciplinary approach will provide just as many benefits to neuroscience as to robotics. The recent explosion of interest in what lies behind psychiatric disorders such as anxiety, depression, and addiction has given rise to a set of sophisticated theories that are complex and difficult to test without some sort of advanced situation platform.
Professor Seymour explained, “We need a way of modelling the human brain to find how it interacts with the world in real-life to test whether and how different abnormalities in these models give rise to certain disorders. For instance, if we could reproduce anxiety behavior or obsessive-compulsive disorder in a robot, we could then predict what we need to do to treat it in humans.”
The team expects that producing robot models of different psychiatric disorders, in a similar way to how researchers use animal models now, will become a key future technology in clinical research.
The team also stated that there may also be other benefits to humans and intelligent robots learning, acting, and behaving in the same way. In future societies in which humans and robots live and work amongst each other, the ability to cooperate and empathize with robots might be much greater if we feel they think like us.
Professor Seymour said, “We might think that having robots with the human traits of being a bit impulsive or overcautious would be a detriment, but these traits are an unavoidable by-product of human-like intelligence. And it turns out that this is helping us to understand human behavior as human.”
The framework for achieving this brain-inspired artificial intelligence was published in two journals, Science Robotics (10.1126/scirobotics.aav2975) on January 16 and Current Opinion in Behavioral Sciences (10.1016/j.cobeha.2018.12.012) on February 6, 2019.
Figure 1. Overview of neuroscience - robotics approach for decision-making. The figure details key areas for interdisciplinary study (Current Opinion in Behavioral Sciences)
Figure 2. Brain-inspired solutions to robot learning. Neuroscientific views on various aspects of learning and cognition converge and create a new idea called prefrontal metacontrol, which can inspire researchers to design learning agents that can address various key challenges in robotics such as performance-efficiency-speed, cooperation-competition, and exploration-exploitation trade-offs (Science Robotics)
2019.02.20 View 6505 -
AI-based Digital Watermarking to Beat Fake News
(from left: PhD candidates Ji-Hyeon Kang, Seungmin Mun, Sangkeun Ji and Professor Heung-Kyu Lee)
The illegal use of images has been a prevalent issue along with the rise of distributing fake news, which all create social and economic problems. Here, a KAIST team succeeded in embedding and detecting digital watermarks based on deep neural learning artificial intelligence, which adaptively responds to a variety of attack types, such as removing watermarks and hacking. Their research shows that this technology reached a level of reliability for technology commercialization.
Conventional watermarking technologies show limitations in terms of practicality, technology scalability, and usefulness because they require a predetermined set of conditions, such as the attack type and intensity. They are designed and implemented in a way to satisfy specific conditions.
In addition to those limitations, the technology itself is vulnerable to security issues because upgraded hacking technologies are constantly emerging, such as watermark removal, copying, and substitution.
Professor Heung-Kyu Lee from the School of Computing and his team provided a web service that responds to new attacks through deep neural learning artificial intelligence. It also serves as a two-dimensional image watermarking technique based on neural networks with high security derived from the nonlinear characteristics of artificial neural networks. To protect images from varying viewpoints, the service offers a depth-image-based rendering (DIBR) three-dimensional image watermarking technique.
Lastly, they provided a stereoscopic three-dimensional (S3D) image watermarking technique that minimizes visual fatigue due to the embedded watermarks. Their two-dimensional image watermarking technology is the first of its kind to be based upon artificial neural works. It acquires robustness through educating the artificial neural networking on various attack scenarios.
At the same time, the team has greatly improved on existing security vulnerabilities by acquiring high security against watermark hacking through the deep structure of artificial neural networks. They have also developed a watermarking technique embedded whenever needed to provide proof during possible disagreements.
Users can upload their images to the web service and insert the watermarks. When necessary, they can detect the watermarks for proof in any dispute.
Moreover, this technology provides services, including simulation tools, watermark adjustment, and image quality comparisons before and after the watermark is embedded.
This study maximized the usefulness of watermarking technology by facilitating additional editing and demonstrating robustness against hacking.
Hence, this technology can be applied in a variety of contents for certification, authentication, distinction tracking, and copyrights. It can contribute to spurring the content industry and promoting a digital society by reducing the socio-economic losses caused by the use of various illegal image materials in the future.
Professor Lee said, “Disputes related to images are now beyond the conventional realm of copyrights. Recently, their interest has rapidly expanded due to the issues of authentication, certification, integrity inspection, and distribution tracking because of the fake video problem. We will lead digital watermarking research that can overcome the technical limitations of conventional watermarking techniques.”
This technology has only been conducted in labs thus far, but it is now open to the public after years of study. His team has been conducting a test run on the webpage (click).Moving forward from testing the technology under specific lab conditions, it will be applied to a real environment setting where constant changes pervade.
1. Figure. 2D image using the watermarking technique: a) original image b) watermark-embedded image c) signal from the embedded watermark
Figure 2. Result of watermark detection according to the password
Figure 3. Example of a center image using the DIBR 3D image watermarking technique: a) original image b) depth image c) watermark-embedded image d) signal from the embedded watermark
Figure 4. Example of using the S3D image watermarking technique: a) original left image b) original right image c) watermark-embedded left image d) watermark-embedded right image e) signal from the embedded watermark (left) f) signal from the embedded watermark (right)
2018.12.05 View 5610 -
AI ｜QC ITRC Opens at KAIST
(from left: Dean of College of Engineering Jong-Hwan Kim, Director of AI│QC ITRC June-Koo Rhee, Vice President for R&DB Heekyung Park and Director General for Industrial Policy Hong Taek Yong)
Artificial Intelligence｜The Quantum Computing Information Technology Research Center (AI｜QC ITRC) opened at KAIST on October 2.
AI｜QC ITRC, established with government funding, is the first institute specializing in quantum computing. Three universities (Seoul National University, Korea University, and Kyung Hee University), and four corporations, KT, Homomicus, Actusnetworks, and Mirae Tech are jointly participating in the center. Over four years, the institute will receive 3.2 billion KRW of research funds.
Last April, KAIST selected quantum technology as one of its flagship research areas. AI｜QC ITRC will dedicate itself to developing quantum computing technology that provides the computability required for human-level artificial intelligence. It will also foster leaders in related industries by introducing industry-academic educational programs in graduate schools.
QC is receiving a great deal of attention for transcending current digital computers in terms of computability. World-class IT companies like IBM, Google, and Intel and ventures including D-Wave, Rigetti, and IonQ are currently leading the industry and investing heavily in securing source technologies.
Starting from the establishment of the ITRC, KAIST will continue to plan strategies to foster the field of QC. KAIST will carry out two-track strategies; one is to secure source technology of first-generation QC technology, and the other is to focus on basic research that can preoccupy next-generation QC technology.
Professor June-Koo Rhee, the director of AI│QC ITRC said, “I believe that QC will be the imperative technology that enables the realization of the Fourth Industrial Revolution. AIQC ITRC will foster experts required for domestic academia and industries and build a foundation to disseminate the technology to industries.”
Vice President for R&DB Heekyung Park, Director General for Industrial Policy Hong Taek Yong from the Ministry of Science and ICT, Seung Pyo Hong from the Institute for Information & communications Technology Promotion, Head of Technology Strategy Jinhyon Youn from KT, and participating companies attended and celebrated the opening of the AI│QC ITRC.
2018.10.05 View 7587 -
Mathematical Principle behind AI's 'Black Box'
(from left: Professor Jong Chul Ye, PhD candidates Yoseob Han and Eunju Cha)
A KAIST research team identified the geometrical structure of artificial intelligence (AI) and discovered the mathematical principles of highly performing artificial neural networks, which can be applicable in fields such as medical imaging.
Deep neural networks are an exemplary method of implementing deep learning, which is at the core of the AI technology, and have shown explosive growth in recent years. This technique has been used in various fields, such as image and speech recognition as well as image processing.
Despite its excellent performance and usefulness, the exact working principles of deep neural networks has not been well understood, and they often suffer from unexpected results or errors. Hence, there is an increasing social and technical demand for interpretable deep neural network models.
To address these issues, Professor Jong Chul Ye from the Department of Bio & Brain Engineering and his team attempted to find the geometric structure in a higher dimensional space where the structure of the deep neural network can be easily understood. They proposed a general deep learning framework, called deep convolutional framelets, to understand the mathematical principle of a deep neural network in terms of the mathematical tools in Harmonic analysis.
As a result, it was found that deep neural networks’ structure appears during the process of decomposition of high dimensionally lifted signal via Hankel matrix, which is a high-dimensional structure formerly studied intensively in the field of signal processing.
In the process of decomposing the lifted signal, two bases categorized as local and non-local basis emerge. The researchers found that non-local and local basis functions play a role in pooling and filtering operation in convolutional neural network, respectively.
Previously, when implementing AI, deep neural networks were usually constructed through empirical trial and errors. The significance of the research lies in the fact that it provides a mathematical understanding on the neural network structure in high dimensional space, which guides users to design an optimized neural network.
They demonstrated improved performance of the deep convolutional framelets’ neural networks in the applications of image denoising, image pixel in painting, and medical image restoration.
Professor Ye said, “Unlike conventional neural networks designed through trial-and-error, our theory shows that neural network structure can be optimized to each desired application and are easily predictable in their effects by exploiting the high dimensional geometry. This technology can be applied to a variety of fields requiring interpretation of the architecture, such as medical imaging.”
This research, led by PhD candidates Yoseob Han and Eunju Cha, was published in the April 26th issue of the SIAM Journal on Imaging Sciences.
Figure 1. The design of deep neural network using mathematical principles
Figure 2. The results of image noise cancelling
Figure 3. The artificial neural network restoration results in the case where 80% of the pixels are lost
2018.09.12 View 6785 -
Center for Industrial Future Strategy Takes Off at KAIST
(Professor Wonjoon Kim from the School of Business and Technology Management)
Professors from KAIST and major international universities launched a mega-scale research center focusing on the Fourth Industrial Revolution, named the Center for Industrial Future Strategy (CIFS).
This center is funded by the National Research Foundation Korea and will receive 2.25 billion KRW over four years.
Directed by Professor Wonjoon Kim from the School of Business and Technology Management, the center is comprised of ten top-tier researchers and four research associates, including Professor Hawoon Jeong (KAIST), Professor Scott Stern (MIT), Professor Aaron Chatterji (Duke University), Dr. Yong Suk Lee (Stanford University) and Professor Hyejin Youn (Northwestern University).
The center will conduct research on technical, social, and economic changes derived by a new paradigm of technological innovation.
Moreover, they will study policies and strategies in relation to innovation in the corporate and government sectors to achieve economic growth in a sustainable manner. The center will also propose policies and strategies in a variety of economic and industrial settings to establish a sustainable and global innovation ecosystem.
To carry out these studies successfully, CIFS will further expand the AIEA-NBER Conference with the Asia Innovation and Entrepreneurship Association (AIEA) and the National Bureau of Economic Research (NBER) in which numerous Nobel Laureates in Economics are affiliated. They will also comprise thematic research teams with co-founding universities to build stronger cooperation with one another.
Besides the academic cooperation, the center will also build partnerships with international organizations, including the Asian Development Bank and the Inter-American Development Bank to carry out their missions at multilateral levels.
Their research topics include changes to value chains in a new paradigm of technological innovation, labor market changes in the Fourth Industrial Revolution, sharing economies and social interests, big data, artificial intelligence & privacy policy, and innovation & ethical and institutional countermeasures to AI technology.
Professor Kim said, “The new paradigm of technological innovation is evolving social, economic, and industrial structures, such as R&D, industry, technology, labor, finance, and institutions. The Center will contribute to proposing policies and strategies so that Korea, as well as the international community, can take appropriate measures to these big changes.”
2018.09.11 View 12029 -
There Won't Be a Singularity: Professor Jerry Kaplan
(Professor Jerry Kaplan gave a lecture titled, Artificial Intelligence: Think Again at KAIST)
“People are so concerned about super intelligence, but the singularity will not happen,” said Professor Jerry Kaplan at Stanford University, an AI guru and Silicon Valley entrepreneur during a lecture at KAIST. He visited KAIST to give a lecture on Artificial Intelligence: Think Again on September 6.
Professor Kaplan said that some people argue that Korea’s AI research is behind the US and China but he doesn’t agree with that. “Korea is the most digitally connected one and has the world’s best engineers in the field. Korean companies are building products the consumers really like at reasonable prices. Those are attracting global consumers,” he added.
Instead of investing loads of money on AI research, he suggested three tasks for Korea taking a better position in the field of AI: Collecting and saving lots of data; training engineers, not the research talents in AI; and investing in AI infrastructure and relieving regulations by the government.
Referring to AI hype, Professor Kaplan argued that machines are intelligent, but they do not think in the way humans can, and assured the audience that the singularity some futurists predict will not be coming. He said, “Machine learning is a tool extracting useful information, but it does not mean they are so smart that they are taking over the world.”
(Professor Jerry Kaplan gave a lecture titled, Artificial Intelligence: Think Again at KAIST)
But what has made us believing AI myths? He first began pointing out how AI has been mythicized by three major drivers. Those are the entertainment industry, the popular media, and the AI community all wanting to attract more public attention and prestige. The abovementioned drivers are falsely making robots more human and are adding human characteristics.
Instead of being captivated by those AI myths and thinking about how to save the world from robots, he strongly argued, “We need to develop standards for the unintended side effects from AI.” To provide machines socially and ethically mingling with the human world, he believed principles should be set as follows: Define the Safe Operating Envelope (SOE), “safe modes” when out of bounds, study human behavior programmatically, certification and licensing standards, limitations on machine “agency,” and basic computational ethics such as when it is okay to break the law.
Professor Kaplan gave a positive view of AI for humans. “The future will be bright, thanks to AI. They do difficult work and help us and that will drive wealth and quality of life. The rich might get richer, but the benefits will spread throughout the people. It is time to think of innovative ways for using AI for building better world,” he concluded.
2018.09.10 View 6079