Papers
1
Total Citations
11
H-Index
1
About
Pedram Pad is a researcher whose work lies at the intersection of deep learning, efficient neural network architectures, and edge computing. His most-cited paper, "Hierarchical Training of Deep Neural Networks Using Early Exiting" (2024, 11 citations), tackles a critical bottleneck in modern AI: the resource-intensive training of deep neural networks (DNNs). Pad proposes a novel hierarchical training framework that leverages early exiting—a technique where intermediate layers of a network can produce predictions—to significantly reduce communication costs, runtime, and privacy risks when training on data from edge devices. This work addresses the growing need for decentralized, efficient AI systems, particularly for vision tasks. By enabling parts of the training to occur locally on edge devices rather than solely on cloud servers, Pad’s contributions help bridge the gap between high-accuracy DNNs and practical deployment in resource-constrained environments. His research is especially relevant for applications in IoT, autonomous systems, and real-time analytics, where latency and data privacy are paramount. With a focus on making deep learning more accessible and efficient, Pedram Pad is shaping the future of distributed and edge-based AI.
Research Focus
Key Achievements
Top Papers
- 1Hierarchical Training of Deep Neural Networks Using Early Exiting11 citations · 2024