每周一起读 × 招募 | ICML 2019:基于粒子的变分推断加速方法
”每周一起讀“是由 PaperWeekly 發(fā)起的論文共讀活動,我們結(jié)合自然語言處理、計算機視覺和機器學習等領(lǐng)域的頂會論文和前沿成果來指定每期論文,并且邀請論文作者來到現(xiàn)場,和大家展開更有價值的延伸討論。
我們希望能為 PaperWeekly 的各位讀者帶來一種全新的論文閱讀體驗、一個認識同好、找到組織的契機、一次與國際頂會論文作者當面交流的機會。
6 月 5 日(周三)晚 7 點半,“每周一起讀”將邀請清華大學計算機系博士生劉暢,和大家分享他發(fā)表于機器學習國際會議 ICML 2019 的兩篇最新文章。
01# 本 期 嘉 賓
? 劉暢??
清華大學計算機系博士生
劉暢,清華大學計算機系博士生,從事統(tǒng)計機器學習方向研究,導師為朱軍教授。他于 2014 年在清華大學物理系取得理學學士學位,博士期間曾在杜克大學訪學一年。他的研究興趣主要在貝葉斯推理方法以及利用幾何結(jié)構(gòu)的機器學習方法。他在機器學習國際會議 ICML, NeurlPS, AAAI 等上發(fā)表了數(shù)篇論文。
? ICML 2019??
Abstract: Particle-based variational inference methods (ParVIs) have gained attention in the Bayesian inference literature, for their capacity to yield flexible and accurate approximations. We explore ParVIs from the perspective of Wasserstein gradient flows, and make both theoretical and practical contributions. We unify various finite-particle approximations that existing ParVIs use, and recognize that the approximation is essentially a compulsory smoothing treatment, in either of two equivalent forms. This novel understanding reveals the assumptions and relations of existing ParVIs, and also inspires new ParVIs. We propose an acceleration framework and a principled bandwidth-selection method for general ParVIs; these are based on the developed theory and leverage the geometry of the Wasserstein space. Experimental results show the improved convergence by the acceleration framework and enhanced sample accuracy by the bandwidth-selection method.
Abstract:?It is known that the Langevin dynamics used in MCMC is the gradient flow of the KL divergence on the Wasserstein space, which helps convergence analysis and inspires recent particle-based variational inference methods (ParVIs). But no more MCMC dynamics is understood in this way. In this work, by developing novel concepts, we propose a theoretical framework that recognizes a general MCMC dynamics as the fiber-gradient Hamiltonian flow on the Wasserstein space of a fiber-Riemannian Poisson manifold. The “conservation + convergence” structure of the flow gives a clear picture on the behavior of general MCMC dynamics. The framework also enables ParVI simulation of MCMC dynamics, which enriches the ParVI family with more efficient dynamics, and also adapts ParVI advantages to MCMCs. We develop two ParVI methods for a particular MCMC dynamics and demonstrate the benefits in experiments.
時間:6 月 5 日(周三) 19:30–21:00
地點:北京智源人工智能研究院6號會議室
北京市海淀區(qū)中關(guān)村南大街1-1號?
中關(guān)村領(lǐng)創(chuàng)空間(信息谷)
長按識別二維碼,即刻報名?
報名截止日期:6?月 5 日(周三)12:00
* 場地人數(shù)有限,報名成功的讀者將收到包含電子門票二維碼的短信通知,請留意查收。
注意事項:
*?如您無法按時到場參與活動,請于活動開始前 24 小時在 PaperWeekly 微信公眾號后臺留言告知,留言格式為放棄報名 + 報名電話;無故缺席者,將不再享有后續(xù)活動的報名資格。
?1 / 掃碼關(guān)注?
掃碼關(guān)注 PaperWeekly?
PaperWeekly
清華大學計算機科學與技術(shù)系
北京智源人工智能研究院
?
現(xiàn)在,在「知乎」也能找到我們了
進入知乎首頁搜索「PaperWeekly」
點擊「關(guān)注」訂閱我們的專欄吧
關(guān)于PaperWeekly
PaperWeekly 是一個推薦、解讀、討論、報道人工智能前沿論文成果的學術(shù)平臺。如果你研究或從事 AI 領(lǐng)域,歡迎在公眾號后臺點擊「交流群」,小助手將把你帶入 PaperWeekly 的交流群里。
▽ 點擊 |?閱讀原文?| 立刻報名
總結(jié)
以上是生活随笔為你收集整理的每周一起读 × 招募 | ICML 2019:基于粒子的变分推断加速方法的全部內(nèi)容,希望文章能夠幫你解決所遇到的問題。
- 上一篇: ACL 2019 | 基于知识增强的语言
- 下一篇: 岗位推荐 | 微软AI Researc