Shun'ichi Amari

{{short description|Japanese scholar (born 1936)}}

{{Infobox scientist

| name = Shun'ichi Amari

| native_name = 甘利 俊一

| native_name_lang = ja

| image = Shunichi Amari cropped Shunichi Amari 201911.jpg

| image_size = 250px

| image_upright =

| alt =

| caption = Shun'ichi Amari

| birth_date = {{birth year and age |1936}}

| birth_place = Tokyo, Japan

| alma_mater = University of Tokyo

| doctoral_advisor =

| academic_advisors =

| doctoral_students =

| notable_students =

| fields = Neuroscience

| known_for = Information geometry
Amari distance
Artificial intelligence

| awards = IEEE Emanuel R. Piore Award (1997)

| workplaces = RIKEN

}}

{{nihongo|Shun'ichi Amari|甘利 俊一|Amari Shun'ichi}}, is a Japanese engineer and neuroscientist born in 1936 in Tokyo, Japan. He was a pioneer in information geometry and artificial intelligence.

Education

He majored in Mathematical Engineering in 1958 from the University of Tokyo then graduated in 1963 from the Graduate School of the University of Tokyo.

His Master of Engineering in 1960 was entitled Topological and Information-Theoretical Foundation of Diakoptics and Codiakoptics.

His Doctor of Engineering in 1963 was entitled Diakoptics of Information Spaces.

Career

Shun'ichi Amari received several awards and is a visiting professor of various universities.

He is the author of more than 200 peer-reviewed articles.[http://www.brain.riken.go.jp/labs/mns/amari/Publications.html Amari Shun'ichi's publication page at RIKEN Brain Science Institute] He is known for developing information geometry. He also independently invented the Hopfield network in 1972,{{cite journal |last1=Amari |first1=Shun-Ichi |date=1972 |title=Learning patterns and pattern sequences by self-organizing nets of threshold elements |journal=IEEE Transactions |volume=C |issue=21 |pages=1197–1206}} a form of self-organized recurrent neural network.

= Artificial intelligence =

Amari made significant advances in artificial intelligence. In 1967, he proposed the first deep learning artificial neural network (ANN) using the stochastic gradient descent (SGD) algorithm. The same year, Amari and his student H. Saito reported the first multilayer perceptron (MLP) neural network trained by SGD.{{cite arXiv |eprint=2212.11279 |class=cs.NE |first=Juergen |last=Schmidhuber |author-link=Juergen Schmidhuber |title=Annotated History of Modern AI and Deep Learning |date=2022}} The concept of backpropagation was also anticipated by Amari in the 1960s.{{Cite web |title=Japanese scientists were pioneers of AI, yet they’re being written out of its history |url=https://www.durham.ac.uk/research/current/thought-leadership/2024/11/japanese-scientists-were-pioneers-of-ai-yet-theyre-being-written-out-of-its-history/ |access-date=2025-07-14 |website=Durham University}}

In 1972, Amari and Kaoru Nakano published the first papers on deep learning recurrent neural networks (RNN). The same year, Amari invented the Amari–Hopfield network. The Amari network, the earliest deep learning recurrent neural network (RNN), was first published by Amari in 1972. It was rediscovered by John Hopfield in 1982 as the Hopfield network.

= Later career =

He is currently holding a position of the RIKEN lab and is vice-president of Brain Science Institute, director of Brain Style Information Systems Group and team leader of Mathematical Neuroscience Laboratory.[http://www.brain.riken.go.jp/labs/mns/amari/Personal.html Amari Shun'ichi's curriculum page at RIKEN Brain Science Institute]

He was a winner of the IEEE Emanuel R. Piore Award (1997).{{cite web |url=http://www.ieee.org/documents/piore_rl.pdf |title=IEEE Emanuel R. Piore Award Recipients |publisher=IEEE |access-date=December 30, 2010 |url-status=dead |archive-url=https://web.archive.org/web/20101124232834/http://ieee.org/documents/piore_rl.pdf |archive-date=November 24, 2010 }}

Key works

  • A Geometrical Theory of Information (in Japanese), Kyoritsu, 1968
  • Information Theory (in Japanese), Daiamondo-sha, 1971
  • Mathematical Theory of Nerve Nets (in Japanese), Sangyotosho, 1978
  • Methods of Information Geometry,[https://books.google.com/books?id=XHGo0sEvr-AC Methods of Information Geometry, Oxford University Press 2000] in collaboration with Hiroshi Nagaoka, originally published in Japanese in 1993 and published in English in 2000 with the American Mathematical Society (AMS).

Awards and honors

References

{{reflist}}