Research
My research centers on Edge AI, aiming to fulfill various user demands on edge devices through AI/ML services. Specific topics include, but are not limited to, the following:
Distributed/Federated Learning
Personalization and Generalization for Trustworthy AI
Data/Resource-Efficient Learning
Based on these thrusts, I have been publishing papers in top-tier ML venues (NeurIPS, ICML, ICLR) as well as top-tier network/communication venues (JSAC, TWC, INFOCOM). Detailed descriptions on each research topic are provided below.
1. Distributed/Federated Learning
With the explosive growth in the number of smart phones and IoT devices, a large portion of data generated nowadays is collected at distributed users at the edge. This strongly motivates the need for distributed/federated learning solutions for training models across the distributed edge network. In the current era of foundation models, there is also a growing demand to fine-tune large language models (LLMs) across distributed clients.
[NeurIPS'23] StableFDG: Style and Attention Based Learning for Federated Domain Generalization
[NeurIPS'21] Few-Round Learning for Federated Learning
[NeurIPS'21] Sageflow: Robust Federated Learning against Both Stragglers and Adversaries
[NeurIPS'20] Election Coding for Distributed Learning: Protecting SignSGD against Byzantine Attacks
2. Personalization and Generalization for Trustworthy AI
Providing trustworthy AI services to every edge user is one of the most rewarding challenges in the current AI/ML era. In particular, models should be personalized to individual users while also effectively handling out-of-distribution scenarios (i.e., generalization/robustness to unseen domains and adversarial attacks). One of my key directions is to utilize foundation models (e.g., LLMs) to effectively address these challenges by simultaneously capturing personalization and generalization capabilities.
[AAAI'24] Consistency-Guided Temperature Scaling using Style and Content Information for Out-of-Domain Calibration
[NeurIPS'23] NEO-KD: Knowledge-Distillation-Based Adversarial Training for Robust Multi-Exit Neural Networks
[INFOCOM'23] SplitGP: Achieving Both Generalization and Personalization in Federated Learning
[ICML'23] Test-Time Style Shifting: Handling Arbitrary Styles in Domain Generalization
3. Data/Resource-Efficient Learning
Training and inference should be conducted in an efficient way to enable AI/ML operations at data/resource-constrained edge devices. Specifically, each user should be able to handle issues related to limited data, limited labels, and limited communication/computation resources during training. In this direction, my goal is to develop efficient algorithms (e.g., lightweight fine-tuning, model/gradient compression, few-shot/active learning) to facilitate learning and inference with large-scale models.
[ICML'24] Achieving Lossless Gradient Sparsification via Mapping to Alternative Space in Federated Learning
[ICLR'23] Warping the Space: Weight Space Rotation for Class-Incremental Few-Shot Learning
[ICLR'23] Active Learning for Object Detection with Evidential Deep Learning and Hierarchical Uncertainty Aggregation
Finally, I have also been actively exploring the above aspects of AI/ML within communication/wireless networks (e.g., 6G satellite networks):
[JSAC'24] Cooperative Federated Learning over Ground-to-Satellite Integrated Networks: Joint Local Computation and Data Offloading
[TMC'24] Federated Split Learning with Joint Personalization-Generalization for Inference-Stage Optimization in Wireless Edge Networks
[JSAC'21] FedMes: Speeding Up Federated Learning with Multiple Edge Servers
[INFOCOM'21] TiBroco: A Fast and Secure Distributed Learning Framework for Tiered Wireless Edge Networks
[TWC'21] Hierarchical Broadcast Coding: Expediting Distributed Learning at the Wireless Edge