Sho Sonoda


Sho Sonoda is a Permanent Senior Research Scientist at Deep Learning Theory Team (PI: Prof. Taiji Suzuki), RIKEN AIP. He received the degree of Doctor of Engineering from Waseda University in 2017 under the supervision of Prof. Noboru Murata. He joined RIKEN in 2018, and he was tenured in 2021. His expertise is in theory and application of machine learning, and especially in harmonic analysis for neural network.

Overview:

Approximation Theory for Deep Neural Networks (MLSS2024 long version): slide v. Mar 2024 (english) 

AIP Open Seminar recordings, Apr2021 (YouTube video, english)

積分表現でニューラルネットを理解する（OCAMI「人工知能と数学」）YouTube動画, slide v2024年3月（日本語）

Specifics:

General Ridgelet Transform Induced from Group Invariant Functions, and Deep Ridgelet Transform: slide v. Dec2023 (english).

Ridgelet Transform for Group Convolutional Neural Networks: slide v. Nov2022 (english) v. 2022年9月（日本語）, poster

Ridgelet Transform for Neural Networks on Noncompact Symmetric Space and Helgason-Fourier Analysis: slide v. Aug2022 (english), poster

Ghosts (or Null Space) in Neural Networks: slide v. June2021 (english)

Approximation Lower Bounds for Neural Networks: slide v. Sep2020 (english) poster v. Jul2023 (english)

Transport analysis for denoising autoencoders: slide v. Aug2019 (english)

Quantum algorithm for optimal random Fourier features: slide v. Nov2021 (english)

積分表現理論と輸送理論（ICML2019読み会資料） : slide v2019年8月 (日本語)

ニューラルネットの零空間（勉強会資料）：slide v2021年10月（日本語）

Research Interest

Contact

sho.sonoda [at] riken.jp

sho.sonoda [at] aoni.waseda.jp (expired at the end of April 2018)

s.sonoda0110 [at] toki.waseda.jp (permanently valid alumni address)