loaction
연구성과
SNU Professor Namkyoo Park's Team Succeeds in Designing Hardware Simulating Brain Characteristics Using Artificial Intelligence
-
작성자
관리자
-
등록일
2020.10.28.
-
조회수
990
SNU Professor Namkyoo Park's Team Succeeds in Designing Hardware Simulating Brain Characteristics Using Artificial Intelligence
- Strong against error, excellent computational efficiency and similar to
signal transmission system efficiency in the brain.
- Development of a basic platform for optical artificial intelligence, published in "Nature Communications"
▲ From left SNU Department of Electrical and Computer Engineering Professor Namykyoo Park, Professor Sunkyu Yu, Doctor Hyunhee Park
SNU College of Engineering (Dean Kookheon Char) announced that a research team led by Professor Namkyoo Park and Professor Sunkyu Yu of the Department of Electrical and Computer Engineering succeeded in designing a hardware with network characteristics similar to the brain by utilizing artificial neural networks of artificial intelligence technology. Similar to the structural characteristics of the brain, the system was found to be capable of very effective computation/signaling of waves like light, quantum, etc. and at the same time was strong against error.
The human brain is a very complex network structure in which hundreds of billions of neurons are connected by thousands of synapses each. The connection structure of these brain neural networks is known to exist in an intermediate area that is neither completely regular nor completely chaotic and is known to be characterized as a 'scale-free network'.
Scale-free networks have unequal structural characteristics that only a few hubs are particularly sensitive to change, so they are strong against general errors, but are fast for signal transfer from within and easy for intentional control. These features are used to explain why the signaling system within the brain is efficient, and it can be assumed from this that its internal structure will naturally have an advantage as a scaleless network when developing artificial intelligence hardware simulating the brain. However, among the numerous candidates for the system, finding a scaleless system that is similar to the brain has been a very challenging task.
Professor Namkyoo Park and Sunkyu Yu's research team have demonstrated that deep learning artificial neural networks can be utilized to identify hardware systems that have similar brain-like characteristics, i.e., characteristics of a scaleless network for waves.
Professor Namkyoo Park, the corresponding author of the paper said, "As a result of acquiring the deep learning neural network that can interpret the interaction between waves and systems, we found that the acquired network provides a scaleless network that is characteristically similar to the brain. When the medium was reverse-designed using this network, it was also highly interesting to note that the scaleless nature of the neural network is projectioned as a structural feature of the hardware system. This means that it is possible to design and organize "brain-like hardware" with "brain-like software (deep learning)". In other words, it is possible to design systems on the implanted neuromorphic hardware using deep learning neuromorphic software."
In addition, "One of the long-standing goals of optical researchers is to implement a computational system that is operated by light. With the current research team planning to publish their "Research/review on disorderly medium" in <Nature Review Materials> in conjunction to the "light-operated neuron related research" published last year in <Advanced Science>, I wish to achieve the development of light-operated artificial intelligence computers that have the advantages of neuromorphic device as well as complex systems. Ultimately, the long-term goal is to present a new methodology for the research of high-speed, light-powered artificial intelligence (Photonic Brain), which has recently been intensively developed by MIT, Stanford, and various startups, as well as to develop the Photonic Quantum Brain based on quantum waves."
Professor Sunkyu Yu, the first author of the paper, said, "The hardware system without a developed scale showed characteristics that are strong against random errors and very sensitive to intended modulation, similar to the characteristics of a network without scale. In other words, it can be applied to efficient switches, logical devices, memory, optical deep learning systems, etc. as we can easily control the behavior of waves by controlling hub neurons while minimizing the impact of mistakes in the process or experimentation, or noise in signal processing during future operation."
This study was conducted with the support of the Global Frontier Project (GFP) of the Ministry of Science and ICT, the Presidential Post-Doc.Fellowship Project (PPD) and the Korea Research Fellowship (KRF), and was published in September 24 of <Nature Communications (IF=12.121)>, a world-renowned international journal.
담당부서대외협력실
전화번호880-9148