High-Dimensional data reduction and acceleration
- Accelerated the construct of knn graph by kd-tree based on LargeVis.
- Proposed an algorithm with sub-linear computation complexity by sharing the gradient between adjacent data points.
Efficient graph layout for large-scale graphs
- Approximating pairwise node distances by a sparse distance matrix and estimating the gradient by using the negative sampling technique.
- Accelerating the optimization process through a multi-level layout scheme.
Visual Analysis of Relationship Network and Behavior Pattern
- Extracted a target individual from heterogeneous data constraints.
- Constructed the daily activity timeline of an individual.
Vision-based pedestrian detection
- Developed algorithms for image feature extraction and feature fusion using MTALAB.
- Used SVM classifier, DS evidence theory to detect the pedestrian in the given image.
National College Students Cloud Computing Application Innovation Contest
- Implemented the algorithm for log analysis based on MapReduce and optimized algorithm performance with Scala.
- Subgraph isomorphism query based on Spark.
Mathematical Contest in Modeling/Interdisciplinary Contest in Modeling
- Leveraged the principal component analysis method to rank the school according to the comprehensive strength.
- Constructed a multivariate linear regression model and used cluster analysis to determine the final schools to be funded.
China Mathematical Contest in Modeling
- Constructed a nonlinear model based on the changes of object’s shadow.
- Exploited the method of single factor control variable to fit the shadow curve, according to the fitting curve and equation to solve out the latitude and longitude.