kaiyun開(kāi)云官方網(wǎng)站

“創(chuàng)源”大講堂研究生學(xué)術(shù)講座

海 報(bào)

講座時(shí)間：2022年12月6日星期二14:00-15:00

講座地點(diǎn)：騰訊會(huì)議：393 272 979；密碼：1206

主講人簡(jiǎn)介：

陳彩華，教授，南京大學(xué)理學(xué)博士，新加坡國(guó)立大學(xué)聯(lián)合培養(yǎng)博士，主持/完成包括國(guó)家自然科學(xué)基金優(yōu)秀青年項(xiàng)目在內(nèi)的多項(xiàng)國(guó)家和省科研項(xiàng)目，代表作發(fā)表于Mathematical Programming，SIAM系列雜志, NeurIPS，CVPR等國(guó)際知名學(xué)術(shù)期刊和會(huì)議。獲華人數(shù)學(xué)家聯(lián)盟最佳論文獎(jiǎng)(2017、2018)，中國(guó)運(yùn)籌學(xué)會(huì)青年科技獎(jiǎng)(2018)，江蘇省工業(yè)與應(yīng)用數(shù)學(xué)學(xué)會(huì)青年獎(jiǎng)(2020),南京大學(xué)青年五四獎(jiǎng)?wù)?span lang="EN-US">(2019)，入選首批南京大學(xué)仲英青年學(xué)者(2017)、南京大學(xué)青年名師名課培育計(jì)劃(2020)及江蘇省社科優(yōu)青(2019)。

講座內(nèi)容簡(jiǎn)介：

Title: Algorithmic Design for Wasserstein DRO Based Trustworthy Machine Learning

In this talk, we consider two essential features of the trustworthy machine learning—fairness and robustness. Mathematical optimization models based on Wasserstein DRO are then proposed to deal with the robust or/and fair machine learning problems. Taking advantage of their specific structure, we design several efficient stochastic algorithms to solve the resulted problems and also establish the corresponding convergence properties. Finally, a few of numerical experiments are presented to illustrating the efficiency of the algorithms.