About me

My research experience includes a research assistantship in Graph and Image Signal Processing Lab (GIPS) at York University with Prof. Gene Cheung, a professor in the EECS department in the Lassonde School of Engineering. I have also worked as a student researcher at Google AI - Perception Lab with Dr. Philip Chou. I have finished my master’s thesis “Volumetric Attribute Compression for 3D Point Clouds using Feedforward Network with Geometric Attention”, which is the cumulative result of my works at both GIPS and Google AI.

I am continuing my research as a Ph.D. student at York. Currently, my main research is about 3D point cloud compression along with mathematically interpretable deep learning models, my latest work, accepted in NISP’24, is “Interpretable Lightweight Transformer via Unrolling of Learned Graph Smoothness Priors” which derived entirely from a graph-based mathematical model, it shares many similarities with attention operations with significantly reduced parameter counts, memory footprint and also gives competitive interpolation performance. Although I have been doing a lot of research work, my desire is to learn more about industry perspectives on machine learning and improve my skill further to a practical level by low-level engineering, e.g, real-time inference, GPU kernel-based implementation.

At the beginning of my career, designing deep learning model architecture was a frustrating process because there is barely any mathematical logic that guides the development (or design) of a deep learning model to tackle a specific problem, it is all about intuition that help and I realized intuition alone is not enough. Hence, I start my research career with this question in my mind, “What are the mathematical seeds that grow into these neural networks architecture? And in obtaining these seeds, would these neural networks architecture become more ‘white-box’”. My belief is that there must be some answer to this question because if more than 50 years of applied mathematics research since the 1970s can give us a glimpse into the universe, then mathematics can not stand useless to these questions.