Jia Li

li.jia AT pku.edu.cn

Assistant Professor

Beijing Normal University


I am a post-doc researcher working with Prof. Zhouchen Lin and Prof. Chao Xu on neural network optimization. I completed my Ph.D. from Beijing Jiaotong University under the supervision of Prof. Zhouchen Lin and Prof. Jian Yu. I received the Excellent Ph.D. Thesis Awards by Beijing Jiaotong University (2017) and Beijing Society of Image and Graphics (2019), respectively.


  • Machine Learning
  • Computer Vision
  • Image Processing
  • Computational Photography


  • Ph.D. in Computer Science, 2012-2017

    Beijing Jiaotong University

  • MSc in Computer Science, 2009-2012

    Zhengzhou University

  • BSc in Mathematics, 2003-2007

    Zhengzhou University

Publications @ZERO Lab

Training Neural Networks by Lifted Proximal Operator Machines. TPAMI, 2020.

We present the lifted proximal operator machine (LPOM) to train fully-connected feed-forward neural networks. LPOM represents the …

Lifted Proximal Operator Machines. AAAI, 2019.

By rewriting the activation function as an equivalent proximal operator, we approximate a feed-forward neural network by adding the …

Demosaicking based on Channel-Correlation Adaptive Dictionary Learning. JEI, 2018.

We propose a channel-correlation adaptive dictionary learning based demosaicking (CADLD).

Automatic Design of High-Sensitivity Color Filter Arrays with Panchromatic Pixels. TIP, 2017.

We propose a fully automatic approach to designing high-sensitivity CFAs using panchromatic pixels based on a mathematical model.

Optimized Color Filter Arrays for Sparse Representation Based Demosaicking. TIP, 2017.

In this paper, we consider optimally designing CFAs for sparse representation-based demosaicking, where the dictionary is well-chosen.

Penrose High Dynamic Range Imaging. JEI, 2017.

We propose the Penrose pixel layout, using pixels in aperiodic rhombus Penrose tiling, for HDR imaging.

Automatic Design of Color Filter Arrays in the Frequency Domain. TIP, 2016.

We present a new method to automatically design CFAs in the frequency domain.

Penrose Demosaicking. TIP, 2015.

Penrose demosaicking is more difficult than regular demosaicking, because none of the color components of the reconstructed regular …