Guest Lecture: Evolutionary Computation for Deep Neural Networks
Date:
This was my second guest lecture for the ECEN5773 - Intelligent Systems course at Oklahoma State University. I introduced undergraduate students to the use of Evolutionary Computation (EC) for designing and optimizing Deep Neural Networks (DNNs), highlighting two major contributions from my own research.
Topics covered:
- Challenges in manual DNN architecture design and pruning
- EC-based approaches for architecture search:
- Particle Swarm Optimization (PSO) for automatic CNN design (psoCNN)
- Evolution Strategy (ES) for filter pruning in CNNs, ResNets, and DenseNets (DeepPruningES)
- Multi-Criteria Decision Making (MCDM) for selecting models based on performance-complexity trade-offs
- Implementation and experimentation on datasets like CIFAR-10
Instructor:
Prof. Gary G. Yen, Regents Professor, OSU