CCICADA Seminar Series in Homeland Security-April 3, 2026
Diversity Preserving Filter Pruning: A Fast Optimization Approach for Practical Neural Network Compression
Date/Time: Friday, April 3, 2026, 1:00 to 2:00 pm Online via Zoom
FEATURED SPEAKER: Weiwei Chen, Rutgers Business School, Rutgers University
Join Zoom Meeting:
https://rutgers.zoom.us/j/97457929898?pwd=LaC6fJxnqJySB7ExYyPRBkhXfIdZAI.1
Abstract:
Modern AI applications increasingly rely on deep neural networks, yet their growing size and computational demands can hinder deployment in resource constrained or real time environments. This challenge appears across domains such as real time recommendation systems, network security, financial fraud monitoring, and intelligent transportation, where timely model updates and fast inference are critical. In these settings, model compression is essential for low latency, efficient use of hardware, and operational reliability. In this talk, we introduce a two stage, diversity driven filter pruning method that delivers reductions in model size while maintaining predictive performance. The key idea is to interpret filter diversity through an information theoretic lens, then prune by keeping a set of filters that are most complementary to one another. We formulate pruning as a p dispersion problem and introduce a polynomial time algorithm with a 2 approximation guarantee relative to the optimal solution, leading to significant pruning time speedups over automated baselines. In numerical experiments, the method achieves substantial pruning time efficiency versus baselines and delivers real wall clock speedups.
Bio:
Weiwei Chen is a Professor in the Department of Supply Chain Management at Rutgers University. His research interests include supply chain optimization, service operations, as well as simulation and global optimization methodologies. He earned his B.S. and M.S. degrees from Tsinghua University and Ph.D. from the University of Wisconsin-Madison. He then worked as a Lead Scientist at GE Global Research Center in New York before joining Rutgers in 2014. He has served as an Associate Editor for INFORMS Journal on Applied Analytics, Service Science, International Journal of Production Research, IEEE Transactions on Automation Science and Engineering, among others. He was Chair of the INFORMS Service Science Section from 2020 to 2023.
Leave a comment