Xuren Zhou
- xurenz@andrew.cmu.edu
- allenchou.github.io
- Pittsburgh
I am a master in systems track of Computational Data Science at Carnegie Mellon University. My interests focuses on storage/cloud system development and motion planning in self-driving systems. I love to learn new technologies and apply them to challenging problems. Currently, I am actively seeking full-time software development engineer opportunities in motion planning and system development.
Work Experiences
Motion Planning Engineer Intern
- Implemented the local planning functionality with traffic rules module design, which improved the maintainability and extensibility of the new planning layer compared to the previous version
- Developed a navigation module with Dijkstra’s algorithm to support planning decision making, which reduced solution space and the lane-change cost of the final decision
- Reconstructed planning map update mechanism with a traversal method to reduce the code complexity
- Refined map assumption library to relieve compatibility issue between map and planning layers
Research Projects
Exploiting Disk-Reliability Heterogeneity in Cluster Storage Systems
- Collaborated with a team of 6 to implement the disk group concept on the top of HDFS, which extended the storage heterogeneity functionality of native HDFS
- Integrated Ganglia system to monitor the cluster under different disk IO metrics
- Benchmarked the performance using DFS-Perf framework to demonstrate the stability of our design
- Improved the flexibility and storage-efficiency of HDFS in the heterogeneous storage cluster
Human Motion Video Enhancement with Visual Effects
- Built an automatic system to superimpose visual effects into videos by template-matching approach, which reduced the editing workload of users
- Implemented STIP detector and HOG3D descriptor using C++ and OpenCV for feature extraction
- Developed dynamic time warping algorithm to match two sequences of video motion features
Video-Driven Facial Expression Animation
- Reconstructed a 3D facial expression model from multi-view marked facial images by BFGS optimization
- Extracted the blendshape weights from the 2D facial video by cascaded regression method
- Built a 3D facial expression animation system by data-driven approach
- Improved the 3D facial expression animation system into real-time by OpenMP parallelization
Quantum Information from Two n-Dimensional Pure Quantum States
Proposed two definitions of anti-parallel spins in n-dimensional Quantum Space and explored the optimal fidelities of n-dimension parallel and anti-parallel spins
Robust Influence Maximization
- Explored the impact of the parameter estimation uncertainty during the influence propagation task
- Designed a greedy algorithm to find an approximate optimal solution with a guaranteed bound
- Conducted two numerical experiments on Flixster and NetHEPT datasets to evaluate our methods
Publications
In this paper, we address the important issue of uncertainty in the edge influence probability estimates for the well studied influence maximization problem — the task of finding k seed nodes in a social network to maximize the influence spread. We propose the problem of robust influence maximization, which maximizes the worst-case ratio between the influence spread of the chosen seed set and the optimal seed set, given the uncertainty of the parameter input. We design an algorithm that solves this problem with a solution-dependent bound. We further study uniform sampling and adaptive sampling methods to effectively reduce the uncertainty on parameters and improve the robustness of the influence maximization task. Our empirical results show that parameter uncertainty may greatly affect influence maximization performance and prior studies that learned influence probabilities could lead to poor performance in robust influence maximization due to relatively large uncertainty in parameter estimates, and information cascade based adaptive sampling method may be an effective way to improve the robustness of influence maximization.