Dr. Eng., Professor (full), WASEDA University, Tokyo, Japan
Department of communications and computer engineering,
(Graduate) school of fundamental science and engineering
Hiroyuki KASAI (HK) is a Professor at Faculty of Science and Engineering, WASEDA University, Tokyo, Japan. He received his B.Eng., M.Eng., and Dr.Eng. degrees in Electronics, Information, and Communication Engineering from WASEDA University in 1996, 1998, and 2000, respectively. He was a research associate at Global Information and Telecommunication Institute (GITI), WASEDA University, during 1998-2002. He was a visiting researcher at British Telecommunication BTexacT Technologies, the U.K., during 2000-2001. He joined Network Laboratories, NTT DoCoMo, Japan, in 2002. He was an associate professor at The University of Electro-Communications (UEC), Tokyo, during 2007-2019, and he was appointed as a professor at UEC in 2019. He was a senior policy researcher at Council for Science, Technology and Innovation Policy (CSTP), Cabinet Office of Japan, during 2011-2013. He was a visiting researcher at Technical University of Munich (TUM), Germany, during 2014-2015. Since September 2019, he has been in his current position.
2000年早稲田大学・大学院理工学研究科・電子情報通信学専攻・博士後期課程修了．大学在学中以来，富永英義・早稲田大学教授（現，名誉教授），亀山渉・早稲田大学助教授（現，教授）に師事．1998年〜2002年早稲田大学・国際情報通信研究センター・助手．この間，2000年〜2001年英国・ブリティッシュテレコム研究所 BTexacT Technologies・訪問研究員．2002年より株式会社 NTT ドコモ・ネットワーク研究所・研究員， 2007年電気通信大学・大学院情報システム学研究科・准教授，2019年電気通信大学・大学院情報理工学研究科・教授．この間，2011年〜2013年内閣府総合科学技術会議（CSTP）事務局・政策統括官付・上席政策調査員（出向），2014年〜2015年ドイツ・ミュンヘン工科大学・訪問研究員． 2019年9月より現職．
My research interests generally include optimization, machine learning and learning-based signal processing with those applications in communication & network systems, image & video processing, and other data analysis fields. Specifically, I am interested in learning and optimization for large-scale structured data and parameters, e.g., non-linear optimization algorithms on Riemannian manifolds, and its applications.
- HK, P.Jawanpuria, and B.Mishra, “Riemannian adaptive stochastic gradient algorithms on matrix manifolds,” ICML, 2019.
- HK and B.Mishra, “Inexact trust-region algorithm on Riemannian manifolds,” NeurIPS (formerly NIPS), 2018.
- HK, H.Sato, and B.Mishra, “Riemannian stochastic recursive gradient algorithm,” ICML, 2018.
- HK and B.Misrha, “Low-rank tensor completion: a Riemannian manifold preconditioning approach,” ICML, 2016.
- HK, “Fast online low-rank tensor subspace tracking by CP decomposition using recursive least squares from incomplete observations,” Neurocomputing, 2019.
- HK, W.Kellerer, and M.Kleinsteuber, “Network volume anomaly detection and identification in large-scale networks based on online time-structured traffic tensor tracking,” IEEE Transactions on Network and Service Management, 2016.
Recent news (since Dec. 2017)
- 06-25-2019 (New !!!)
- “Riemannian optimization on the simplex of positive definite matrices” (B.Mishra, HK, and P.Jawanpuria) has been published in arXiv paper.
- “Riemannian adaptive stochastic gradient algorithms on matrix manifolds” (HK, P.Jawanpuria, and B.Mishra) has been accepted in ICML2019.
- “Riemannian stochastic variance reduced gradient algorithm with retraction and vector transport” (H.Sato, HK, and B.Mishra) has been accepted in SIAM Journal on Optimization.
- “A Riemannian gossip approach to subspace learning on Grassmann manifold” (B.Mishra, HK, P.Jawanpuria, and A.Saroop) has been accepted in Machine Learning.
- “Fast online low-rank tensor subspace tracking by CP decomposition using recursive least squares from incomplete observations” (HK) has been accepted in Neurocomputing.
- “McTorch, a manifold optimization library for deep learning” (M.Meghwanshi, P.Jawanpuria, A.Kunchukuttan, HK, and B.Mishra) has been accepted in NeurIPS workshop MLOSS2018.
- “Stochastic optimization library: SGDLibrary” (HK) has been accepted in NeurIPS workshop MLOSS2018 (spotlight).
- “SimpleDeepNetToolbox (Simple Deep Net Toolbox in MATLAB)” has been released.
- “McTorch (Manifold optimization library for deep learning)” has been released.
- “Fast optimization algorithm for hybrid precoding in Millimeter wave MIMO systems” (HK) has been accepted in GlobalSIP2018.
- “Inexact trust-region algorithm on Riemannian manifolds” (HK and B.Mishra) has been accepted in NeurIPS2018 (formerly NIPS).
- “Stochastic recursive gradient on Riemannian manifolds” (HK, H.Sato, and B.Mishra) has been accepted in ICML workshop GiMLi2018.
- “Low-rank geometric mean metric learning” (M.Bhutani, P.Jawanpuria, HK, and B.Mishra) has been accepted in ICML workshop GiMLi2018.
- “Riemannian joint dimensionality reduction and dictionary learning on symmetric positive definite manifolds” (HK and B.Mishra) has been accepted in EUSIPCO2018.
- “Accelerated stochastic multiplicative update with gradient averaging for nonnegative matrix factorization” (HK) has been accepted in EUSIPCO2018.
- “Riemannian stochastic recursive gradient algorithm” (HK, H.Sato, and B.Mishra) has been accepted in ICML2018.
- “SGDLibrary: A MATLAB library for stochastic optimization algorithms” (HK) has been accepted in JMLR.
- “Stochastic variance reduced multiplicative update for nonnegative matrix factorization” (HK) has been accepted in ICASSP2018.
- “Riemannian stochastic quasi-Newton algorithm with variance reduction and its convergence analysis” (HK, H.Sato, and B.Mishra) has been accepted in AISTATS2018.
Email: hiroyuki.kasai at waseda.jp (Please replace the word “at” with “@”.)