Photo

Sirui Xie

PhD Student
Center for Vision, Cognition, Learning and Autonomy
Computer Science Department, UCLA
Email: srxie [at] ucla [dot] edu

Google Scholar | LinkedIn | Twitter | GitHub

I am a third-year Ph.D. student in Computer Science Department, UCLA, advised by Prof. Song-Chun Zhu. As a member of Center for Vision, Cognition, Learning and Autonomy (VCLA), I am also actively advised by Prof. Ying Nian Wu. I obtained BEng degree in Computer Science from The Hong Kong University of Science and Technology (HKUST). In the summer 2021, I worked as a research intern at Facebook AI Research (FAIR), hosted by Rama Vedantam, Ari Morcos, and Stephane Deny.

"A reasonable goal of the visual system, [...] is to extract statistical dependencies so that images may be explained in terms of a collection of independent events" (Olshausen and Field, 1996). My current research focuses on learning representation functions such that the resulted coding can recover the (conditionally) independent factors of our highly structured world, so as to emerge the cognition of "semantic meanings". A representation that successfully matches such a requirement is expected to exhibit systematicity and generalizability:

  • Being systematic in the sense of being sparsely distributed in a low-dimensional manifold, where the vector coding is equivariant to the transformation for each factor of variation and invariant to other irrelevant factors;
  • Being generalizable in the sense of productively composing/correlating intuitively independent factors to extrapolate beyond training domains, and gracefully preserving causal invariance under distribution shifts.

"True understanding enables predictions in novel situations, where some mechanisms change and others are added" (Pearl 2009). The ultimate goal is to build artificial agents with such representations that they can understand novel contexts, imagine unseen situations, and design causal interventions, all at the abstract level, hence efficiently plan for intelligent strategies. I draw inspiration broadly from classical AI, cognitive science, computational neuroscience, statistical learning, and causality. Previously, I worked on (inverse) reinforcement learning and evolving neural representation functionals (a.k.a. Differentiable Neural Architecture Search).

News

Publications ( show selected | show all )

 

Education

  • 2019.09 - present, University of California, Los Angeles
    PhD in Computer Science
  • 2012.09 - 2016.06, The Hong Kong University of Science and Technology
    BEng in Computer Science, First-Class Honor

Selected Awards

  • Graduate Research Assistantship, UCLA, 2019 - present
  • Full Scholarship, HKUST, 2012 - 2016

Professional Service

  • Conference Reviewer: NeurIPS, ICML, ICLR, AISTATS, AAAI, IJCAI, CVPR, ICCV, ECCV, ICRA
  • Journal Reviewer: IEEE/T-PAMI, IEEE/RA-L

Contact

3878 Slitcher Hall
603 Charles E Young Dr E
Los Angeles, CA 90024
srxie [at] ucla [dot] edu