Basic Info


  •   Leo Xu
  •   Computer Software
      and Theory
  •   Master
  •   Peking University
  •   2019

Contact


  • +86  13120180576
  • LIYOU XU PKU
  • leoxupku@gmail.com
  • github.com/MikeXuPku gitee.com/LeoPkuCs

Skill Point


Education


  • Peking University, Theory and Application of Computer, Master
  • Technion, Machine Learning, Exchange
  • Peking University, Computer Science, Bachelor

Awards


  • National Scholarship

  • School Special Scholarship

  • Full Scholarship in Technion Summer School

  • Outstanding Graduates Prize in Beijing Province

  • Outstanding Graduates Prize in Peking University

  • Carku Scholarship

  • 8085 School Scholarship

  • Robin Li Scholarship

Project & Achievement


  • Physically Based Simulation, Cloth Simulation

    2017~,Physika 2.0,one of the principals
    I'm in charge of cuda implementation of Finite Element Analysis, Peridynamics Analysis, some time integration method like explicit euler, runge-kunta and implicit quasi-newton, nonlinear conjugate gradient. As for the whole project, I maintain the CMakelists, auto-doc scripts(doxygen, sphinx), support the python binding of our library based on pybind11. I am doing some research work about extending peridynamics in graphics simulation and animation. It can be applied on 3D,2D,1D objects uniformly and can be used on cloth simulation.

    2015~2017 Physika 1.0,one of the developers
    I'm in charge of implementation of Finite Element Analysis. Physika 1.0 is physically based simulation library implemented by C++. It is aimed at saving the devolping time for researchers. Homepage:github.com/PhysikaTeam/Physika

  • Machine Learning & Deep Learning

    2018~, reverse estimation of material parameters for deformation,researcher
    infering the material parameters like young's module, poisson ratio according to the input deformation and output force load or deformation energy estimation. It is simply modeled as a linear regression problem at the begining which only has one hidden layer between strain and stress. I implement the basic back propogation, optimization methods like SGD and momentum in Cuda C++ for my experiment network because its training data is deeply coupled with my C++ simulation project. Now I am trying to wrap my C++ simulation library with python and extending the network to abstract training data from deformation video rather than the well prepared data.

    2017~2018, course projects about object detection and image classification based on tensorflow API, student
    doing the classical deep learning exercise like traffic sign detection and recognition, liver medical image classification by tensorflow.
  • Web Developing

    2017, Online booking system 2.0, full stack
    I reconstructed the website by nodejs+Angular+Meteor. For the UI component, I used the NG-ZORRO which is UI material libaray provided by Alibaba Ant Finance.

    2016~2017,Online booking system for laboratory instruments,full stack
    I was entrusted by the Geoscience school in Peking University to provide an online resolution for management of laboratory instruments. Front: html5, jquery, bootstrap, velocity Back: java, springmvc, postgresql, mybatis

    2015~2016, www.bedancer.cn,back-end developer
    Bedancer is a website serving dancing comunity. Users can register any dance competition or activity and pay the fee online. I'm in charge of the design of database, implementation of API service.

Activities


  • President of association, sports dance association of Peking University, Feb 2016 ~ Feb 2017

  • Class monitor, class 3 of EECS12, Sep 2013 ~ Jul 2016

Paper


Xu, L., He, X., Chen, W., Li, S., Wang, G. Reformulating Hyperelastic Materials with Peridynamic Modeling. (accepted by Pacific Graphics 2018)

Zhao, J., Zhu, F., Tang, Y., Xu, L., Li, S., Wang, G. Real-Time Example-Based Materials in Laplace-Beltrami Shape Space.