Research
Past and Current Research
This is the Compressed Binary Representation of a graph
Developed and published Bit-packed Compressed Sparse Row (CSR) algorithm for efficient and lossless compression for time-evolving graph data
The compression technique achieved about 50% better compression time and about 30% increase in the compression ratio compared to the current state-of-the-art Compressed Binary Tree (CBT)
Developed and published a structured optimized Matrix-Matrix multiplication on the compressed matrices using partial sum implementation
Developed and published work in billion-scale tensor data, with respect to compression and parallel algorithm
All research is performed in C/C++
Future Research
At present, we are trying to look to model Graph Neural Nets (GNN), to answer some of the graph queries, especially on a time-evolving social networks
In the future, we also want to develop a query based graph compression technique
src: pngWing