Department of Information Engineer, CUHKSep 2020 - Jun 2021
- Researched on scalability of a Graph Neural Network on financial data analytics.
- Implemented a pipeline with PyTorch and DGL for to detect anomaly users/activities. The framework reduced training time by 10-20% and gave close results on real-world business datasets.
- Cooperated with a team of 6 and designed a web-based user portal for graph analysis, which integrated node classification and link prediction objectives using transfer learning and active learning techniques.
Department of Computer Science, CUHKMay 2019 - Sep 2019
- Conducted hotspot detection on layout patterns with several machine learning-based models
- Designed robust loss evaluation methods to deal with outlier data points.
- Implemented a generative neural network that boosted valid VLSI pattern number by 5.8% and pattern library diversity by 21.4%. The framework also enabled context-specific pattern generation with paper accepted.