CV

Yunzhen Feng

No. 5 Yiheyuan Road, Beijing, China.
felix.feng681@gmail.com

Education

• School of Mathematical Sciences, Peking University
  B.S. in Mathematics, 2017 - 2021(expected).
  Major GPA: 3.72/4.0 (Around 30% students cound be higher than 3.58 according to the Dean’s regulation).
  Supported by Elite Undergraduate Training Program of Pure Math and Applied Math.
  

• The High School Affiliated to Renmin University of China
  2011 - 2017

Research Experiences

Research Assistant(2019.02 - Present)
At Beijing International Center for Mathematical Research(BICMR).
Advised by Prof. Bin Dong.
Mainly on adversarial robustness, generalization theory, and NeuralODE. We have proposed a generalization error estimate from Neural ODE perspective and have explained properties of adversarial examples via neural tangent kernel. Currently we are considering the influence on model robustness of injecting different randomness in model and training process.

Other Course Projects

• 2019 Combining traditional energy-based methods with virtual adversarial training in semi-supervised learning:
  We proposed a new perspective to formulate the regularization effects that virtual adversarial training(VAT) impose during training process. In this view point, connection between traditional energy methods and VAT can be bridged and several modification are made to the VAT. We also have further ideas on designing new regularization and it is still an ongoing work.
• 2019 Long-term Traffic Flow Prediction:
  We investigate and implement various influencial models for traffic forecasting and networks for temporal graphs. Based on previous studies, we optimize the networks structure and make modification for 3D-convolution-based and WaveNet-based methods, improving their results on real world dataset.
• 2018 Algorithms for Large-Scale Optimal Transport:
  In this work, we combine Newton Method and Sinkhorn Iteration into a faster and more stable algorithm.

Skills

• Computing Skills: C++, Python (PyTorch), Matlab, LATEX.
• Mathematical Skills: Stochastic Process, Optimal Transport, Differential Geometry.
• Numerical Algorithms: Numerical Methods, Optimization.

Awards

Academic Excellence Award(5%) in 2019,
Meritorious Award for MCM/ICM (top 9%) in 2019,
The elite undergraduate training program of Applied Math 2019-present(top 15%),
The elite undergraduate training program of Pure Math 2018-present,
China Mathematics Olympiad, Gold Medal(same level as USAMO), 2016,
China Mathematics Competition, First Place in Beijing, 2016,
International Space Settlement Design Competition, Finalist, 2016,
NSS Space Settlement Design Contest, Top Prize for Grade 10, 2015.