Projects

Document Analytics directly on Compressed Data (VLDB’18, ICS’18, ICPP’19, ICDE’20) introduction of research projects

• We propose the first known solution to enable direct document analytics on compressed data, which saves 90.8% storage space and 77.5% memory usage, while halving the analytics time.

Intelligent Programming in Heterogeneous Systems (MACSOTS’15, CGO’17, TPDS’17, TJSC’18, TKDE’19, ICPP’20, USENIX ATC’20)

• We develop domain-specific languages, tools, and systems for heterogeneous processors, such as:
• FinePar: Fine-grained partitioning tool for irregular applications
• FineStream: Fine-grained stream processing system

Machine Learn Systems (IJCAI’20, ICPP’20, NPC’19, more papers are on the way)

• We are designing novel deep learning algorithms and systems, such as PewLSTM (periodic LSTM with weather-aware gating mechanism), or ParSecureML (the first secure machine learning framework on GPUs).