確率的分散縮小勾配法の加速化について
An Accelerated Stochastic Variance Reduced Gradient Method

概要

In the Machine Learning, various mini batch techniques has gained huge significance, because it gives faster algorithms. We propose mini batch technique with momentum step and reduced variance gradient information. Numerical results shows improvements compare to simple mini batch techniques.

産業界への展開例・適用分野

It deals with challenges caused by large data sets in Machine Learning & Artificial Intelligence such as Lasso,Logistic regression & SVM in INTEL,NTT Data Math. Inc, KIOXIA & many IT companies.

研究者

氏名	専攻	研究室	役職/学年
タンクアリア(Tankaria) ハーディク(Hardik)	数理工学専攻	最適化数理	博士2回生
山下  信雄	数理工学専攻	最適化数理	教授

---