Panoptic

Panoptic

Kunihiko Fukushima

Sources
ArabicEgyptian ArabicGermanEnglishFrenchJapanese
Kunihiko Fukushima (Japanese: 福島 邦彦, born March 16, 1936) is a Japanese computer scientist and artificial intelligence researcher. Regarding his birthplace, French Wikipedia records that he was born in Saga, Japan, whereas Arabic, English, and Japanese sources indicate he was born in Taiwan (Japanese Taiwan). He is best known as a pioneer in the field of artificial neural networks and deep learning, with work spanning convolutional neural network architecture, unsupervised learning, and the ReLU activation function. Fukushima studied electronics at Kyoto University, earning his bachelor's degree in 1958 and his doctorate in 1966.[6]
He previously worked as a senior research scientist at the NHK Science and Technology Research Laboratories, served as a professor at Osaka University from 1989 to 1999, at the University of Electro-Communications in Chōfu from 1999 to 2001, and at the Tokyo Institute of Technology from 2001 to 2006, followed by a visiting professorship at Kansai University from 2006 to 2010.[7] Currently, he works part-time as a senior research scientist (or special research fellow) at the Fuzzy Logic Systems Institute in Iizuka City, Fukuoka Prefecture, Japan.[7]
Fukushima has been developing neural network models since 1965, specifically to emulate higher brain functions, particularly of the visual system. In 1979, he developed the Neocognitron, a hierarchically structured convolutional artificial neural network that recognizes visual patterns through the application of various learning algorithms and methods, which is considered one of the precursors of modern deep neural networks.[18] One of these methods is the shift of perceptual attention (selective attention model), with which he demonstrated the recognition and completion of occluded objects. Fukushima also developed neural networks for the detection of optical flow, visual motion, the determination of symmetry axes, and other tasks.
He was the founding president of the Japanese Neural Network Society (JNNS) and served on the board of the International Neural Network Society (INNS).[8] His honors include the IEICE Achievement Award and Distinguished/Excellent Paper Awards, the IEEE Neural Networks Pioneer Award (2003), the APNNA Outstanding Achievement Award, the JNNS Distinguished/Excellent Paper Award, the INNS Helmholtz Award, and the Bower Award and Prize for Achievement in Science (2021).[8]

Education and Career

Fukushima studied at Kyoto University, graduating in 1958 with a Bachelor of Engineering in electronics from the Department of Electronic Engineering, Faculty of Engineering.[1,5] He obtained his doctorate in electronics at Kyushu University in 1961. He began his career as a researcher at the Toshiba Research Institute and then joined the Neuro-Informatics Laboratory at Tohoku University. He also served as a Senior Research Scientist at the NHK Broadcasting Science Research Laboratories and the Science & Technology Research Laboratories of the Japan Broadcasting Corporation (NHK).
In 1989, he joined the faculty of Osaka University as a professor at the School of Engineering Science.[1] In 1999, he became a professor at the Faculty of Electro-Communications at the University of Electro-Communications, and in 2001, he joined the faculty of Tokyo University of Technology as a professor. From 2006 to 2010, he was a visiting professor at Kansai University, and since 2006, he has served as a Special Research Fellow at the Fuzzy Logic Systems Institute.[4,17] He is also a professor emeritus at Fukushima University and has participated in numerous international research projects on deep learning and machine learning.
His work focuses primarily on neural networks, computer vision, and cognitive models inspired by the brain; in the 1980s, he developed the Neocognitron, a neural network model capable of recognizing visual patterns in a hierarchical manner and robust against deformations. This model constitutes a major theoretical foundation for modern convolutional network architectures used in image recognition and artificial intelligence. Fukushima served as the first and founding president and an honorary member of the Japanese Neural Network Society (JNNS), and was named a Fellow of the Institute of Electronics, Information and Communication Engineers (IEICE). He was also a founding member of the Board of Governors of the International Neural Network Society (INNS), serving on its board from 1989 to 1990 and from 1993 to 2005, and served as president of the Asia Pacific Neural Network Assembly (APNNA).[1,2] In 2021, he received the Bower Award and the C&C Prize. He has also received the IEICE Achievement Award, the IEICE Best Paper Award, the IEEE Neural Networks Pioneer Award, the APNNA Excellent Paper Award, the JNNS Excellent Paper Award, and the INNS Helmholtz Award.[19,2]

Scientific Achievements

In 1980, Fukushima published the Neocognitron, the original prototype model of a deep convolutional neural network (CNN) architecture.[2,9,10,11,20] Fukushima proposed several supervised and unsupervised learning algorithms to train the parameters of the deep Neocognitron so that it could learn internal representations of incoming input data.[2,3,12] Today, however, convolutional neural network architectures are usually trained through backpropagation. This approach is currently widely and heavily used in the field of computer vision.[2,4,12,13]
In 1969, Fukushima introduced the ReLU (Rectified / Rectifier Linear Unit) activation function in the context of visual feature extraction in hierarchical neural networks, which he called the "analog threshold element".[5,4,3] Although the ReLU function was first used by Alston Householder in 1941 as a mathematical abstraction of biological neural networks, it has been regarded as the most popular activation function for deep neural networks as of 2017.[2,16,15,14]

Impact and Legacy

Fukushima's work has had a major influence on the development of convolutional neural networks (CNN), which are used today in image recognition, video, machine translation, and other fields of artificial intelligence. His concepts of hierarchy and multi-layer processing were adopted and extended by researchers such as Yann LeCun and other pioneers of deep learning.

Awards and Distinctions

In 2020, Fukushima received the Bower Award and Prize for Achievement in Science.[17] In 2022, he was named a laureate of the Asian Scientist 100 by Asian Scientist magazine. He also received the Institute of Electronics, Information and Communication Engineers (IEICE) Achievement Award and Excellent Paper Awards, the IEEE Neural Networks Pioneer Award, the Asian Pacific Neural Network Society (APNNA) Outstanding Achievement Award, the Japanese Neural Network Society (JNNS) Excellent Paper Award, and the International Neural Network Society (INNS) Helmholtz Award.[4,5]

External Links

External links include a ResearchMap profile.
Name
Kunihiko Fukushima
Born
March 16, 1936
Birthplace
Japanese Taiwan
Citizenship
Japan
Fields
Computer science
Workplaces
Fuzzy Logic Systems Institute
Alma mater
Kyoto University
Known for
Artificial neural networks, Neocognitron, Convolutional neural network architecture, Unsupervised learning, Deep learning, ReLU activation function
Awards
IEICE Achievement Award and Distinguished Paper Awards, IEEE Neural Network Pioneer Award, APNNA Outstanding Achievement Award, JNNS Distinguished Paper Award, INNS Helmholtz Award, Bower Award and Prize for Achievement in Science
Sources
العربيةمصرىDeutschEnglishFrançais日本語

See Also

References

  1. [1]
    ^ ieeetv.ieee.org[Arabic]
  2. [2]
    ^ [Arabic]
  3. [3]
  4. [4]
    ^ www.youtube.com[Arabic]
  5. [5]
    ^ hal.science[Arabic]
  6. [6]
  7. [7]
    ^ researchmap.jp[German]
  8. [8]
    ^ www.uec.ac.jp[German]
  9. [9]
    ^ A theory of steady-state activity in nerve-fiber networks: I. Definitions and preliminary lemmasThe Bulletin of Mathematical Biophysics by Alston S. Householder (June 1941)[English]
  10. [10]
    ^ Searching for Activation Functions by Prajit Ramachandran; Zoph Barret; V. Le Quoc (October 16, 2017)[English]
  11. [11]
    ^ A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in positionBiological Cybernetics by Neocognitron Fukushima[English]
  12. [12]
    ^ NeocognitronScholarpedia by K. Fukushima[English]
  13. [13]
    ^ A History of Deep Learning by Andrew Fogg[English]
  14. [14]
    ^ Visual feature extraction by a multilayered network of analog threshold elementsIEEE Transactions on Systems Science and Cybernetics by K. Fukushima (1969)[English]
  15. [15]
    ^ Competition and Cooperation in Neural NetsSpringer by K. Fukushima; S. Miyake (1982)[English]
  16. [16]
  17. [17]
    ^ Kunihiko Fukushima (2020-01-25)[English]
  18. [18]
    ^ doi.org[French]
  19. [19]
  20. [20]
    ^ Deep learningNature by Yann LeCun; Yoshua Bengio; Geoffrey Hinton[Japanese]

External Links

Article Statistics

Word Count Comparison

Comparing content volume across 6 language sources

PanopticPanopticAggregated
1,083 words
Unique (1 source)Full consensus (6 sources)
Englishen
825 words
Arabic(العربية)ar
776 words
Japanese(日本語)ja
773 words
German(Deutsch)de
301 words
French(Français)fr
287 words
Egyptian Arabic(مصرى)arz
30 words
6
Language Sources
1,083
Aggregated Words
138
Full Consensus
369
Unique Claims
1
Disagreements