学术交流

Machine Learning Theory Workshop

TIME: June 20-21, 2019

LOCATION: Lecture Hall, Jiayibing Building, Jingchunyuan 82, Peking University

 

Conference Introduction/

Description

Over the last few years, tremendous progress has been made in machine learning, in terms of both applications and algorithms. At the same time, there also have been a number of novel works on theoretical understandings for different learning problems and algorithms, but many challenges still remain. The goal of this workshop is to study the challenges in machine learning theory, especially to foster discussion and spur research between experts and researchers in the field that can solve fundamental problems, as well as to promote academic communication and cooperation between faculties and students from Peking University and overseas outstanding scholars. 


Scientific Committee 

Weinan E (Princeton University)

Wen Gao (Peking University)

Pingwen Zhang (Peking University)

 

Organization Committee 

Bin Dong (Peking University)

Liwei Wang (Peking University)

Zhihua Zhang (Peking University)

 

Invited Speaker

Stephen Boyd (Stanford University)

Simon Du (Princeton University)

Quanquan Gu (University of California, Los Angeles)

Wei Hu (Princeton University)

Nan Jiang (University of Illinois at Urbana-Champaign)

Qianxiao Li (Institute of High Performance Computing, Singapore)

Ruoyu Sun (University of Illinois at Urbana-Champaign)

Yuhao Wang (Massachusetts Institute of Technology)

Huan Xu (Georgia Institute of Technology)

Qi Yu (Northeastern University)

 

Schedule

               

                   2nd Machine Learning Theory Workshop

 

June 20

Title

Speaker

9:00-10:00

Keynote

TBA

Stephen Boyd 

(Stanford)

Break

10:15-11:00

Sample-efficient exploration in reinforcement learning with function approximation

Nan Jiang

(UIUC)

11:00-11:45

Continuous-time methods for stochastic optimization and batch normalization

Qianxiao Li

(HPC)

11:45-1:30

Lunch

 

1:30-2:15

Physics Guided Deep Learning for Large-Scale Spatiotemporal Analysis

Rose Yu 

(NEU)

2:15-3:00

Towards Understanding Overparameterized Deep Neural Networks: From Optimization To Generalization

Quanquan Gu

(UCLA)

Break

3:15-4:00

Learning High-Dimensional Graphical Models under Total Positivity without Tuning Parameters

Yuhao Wang

(MIT)

4:00-4:45

On the connection between over-parameterized neural networks and kernels, and how to make that useful

Wei Hu

(Princeton)

Break

5:00-5:20

Appropriate function spaces for two-layer neural network and residual network models

Chao Ma

(Princeton)

5:20-5:40

A Regularized Approach to Sparse Optimal Policy in Reinforcement Learning

Xiang Li

(PKU student)

5:40-6:00

Distributed Bandit Learning: Near-Optimal Regret with Efficient Communication

Yuanhao Wang

(THU student)

6:30

Banquet

 

June 21

Title

Speaker

9:00-9:45

Title: The online saddle point problem and its applications

Huan Xu

(Georgia Tech)

9:45-10:30

Towards Deeper Understanding of Neural Network Landscape: Local Minima and Decreasing Paths

Ruoyu Sun

(UIUC)

Break

10:45-11:30

Provably Efficient Reinforcement Learning with Function Approximation

Simon Du

(Princeton)

11:30-11:50

A Priori Estimates of Generalization Error for Two-layer and Residual Networks

Lei Wu

(Princeton)

11:50-12:10

A Gram-Gauss-Newton Method Learning Over-parameterized Deep Neural Networks for Regression Problems

Tianle Cai

(PKU student)

12:10-2:00

Lunch

 

 

Accommodation and Travel Information

Invited speakers will stay in No.9 building of "Zhong Guan Xin Yuan" Hotel. By default, rooms have been booked for the arrival on June 19 and departure on June 22. The hotel is across the street of Peking University Campus, about 15-20 minutes' walk to the Lecture Hall.

PKU will cover travel expenses for all the speakers.
Other participants are responsible for arranging and paying for their own reservations for hotel accommodations and travel expenses. If anyone needs help, please contact Ms. Yanyun Liu.

Hotel Address: “Zhong Guan Xin Yuan” Hotel (Zhongguanyuan Global Village) 
No. 126 Zhongguancun North Road, Haidian District, Beijing 100871, China 
Tel: + (86 10) 62752288 
Fax: + (86 10) 62752236

From the Airport to the Hotel: The most convenient way to get the hotel from the airport is to take a taxi. It costs about RMB 120 yuans. You are also responsible to pay the toll of 10 yuans. So you need some pocket money. A possible way to get cash is to find an ATM in the lobby of the airport. You may print this slip and show it to the taxi driver.        

Contact

Mailing address: Yanyun Liu

Room 1266, Science Building No .1, Peking University, Beijing, China 100871

Phone: +86 10 62768630

E-mail: yyliu@pku.edu.cn

Fax:+ (86 10) 62751801

 

Tips

Gmail account is NOT accessible in China. If you get used to gmail account, please prepare for an alternative email account. Google, Tweeter, Facebook, Youtube etc are not available.

相关附件:
  • schedule_20190620.docx
  • 学术交流 ACADEMIC EXCHANGE