Research Interests
- Conditional computation and early exit in neural networks
- Model-distributed learning and inference
- Federated unsupervised learning and clustering, decentralized learning
- Learning and communication in unmanned aircraft systems
- Vector quantization with applications to location optimization and node placement
- Network connectivity and percolation
- Limited feedback in wireless multi-antenna communication systems
Highlights
- E. Koyuncu, “Theoretical Guarantees for One-Shot Pruning and Compute-Adaptive Early Exit,” ICML 2026.
- LANMAN 2025 Best Paper Award, “Model-Distributed Inference for Large Language Models at the Edge.”