A5057940216- Safety Warnings and Signage
- Risk and Safety Analysis
- Model Reduction and Neural Networks
- Advanced MIMO Systems Optimization
- Stochastic Gradient Optimization Techniques
- Smart Grid Energy Management
- Privacy-Preserving Technologies in Data
- Reinforcement Learning in Robotics
- Traffic Prediction and Management Techniques
- Domain Adaptation and Few-Shot Learning
- Generative Adversarial Networks and Image Synthesis
Arizona State University
2023-2024
A central question in edge intelligence is "how can an device learn its local model with limited data and constrained computing capacity?" In this study, we explore the approach where a global initialization first obtained by running federated learning (FL) across multiple devices, based on which semi-supervised algorithm devised for single to carry out quick adaptation data. Specifically, account heterogeneity resource constraints, trained via FL, each conducts updates only customized...
Headway, namely the distance between vehicles, is a key design factor for ensuring safe operation of autonomous driving systems.There have been studies on headway optimization based speeds leading and trailing assuming perfect sensing capabilities.In practical scenarios, however, errors are inevitable, calling more robust to mitigate risk collision.Undoubtedly, augmenting safety would reduce traffic throughput, highlighting need incorporate both tolerance models.In addition, prioritizing...
Learning generative models is challenging for a network edge node with limited data and computing power. Since tasks in similar environments share model similarity, it plausible to leverage pretrained from other nodes. Appealing optimal transport theory tailored toward Wasserstein-1 adversarial networks (WGANs), this study aims develop framework that systematically optimizes continual learning of using local at the while exploiting adaptive coalescence models. Specifically, by treating...
Real-time traffic has stringent requirements in terms of latency, and deadline guarantees on packet delivery play a vital role real-time IoT applications. Deadline-aware wireless scheduling been long-standing open problem, despite significant efforts using analytical methods. Departing from the conventional approaches, this work studies deadline-aware by taking an offline reinforcement learning (RL) approach to train algorithms, ready be used for online scheduling. To address challenges...