Research Areas
I work on efficient and explainable artificial intelligence (XAI), the wireless world model (WWM) — our current research focus — and learning-based optimization (L2O) for real-world communication and network decisions. Our goal is interpretable, editable, and deployable methods that connect physics, data, and decision making.
Current Focus: Wireless World Model (WWM) 3DGSNeRF
Physically grounded, data-driven models of radio EM environments that fuse geometry and materials (e.g., BIM, point clouds, maps), contextual signals (vision, language, IMU, GNSS), and RF measurements (sweeps, arrays). We combine neural representations (3DGS and NeRF variants, diffusion and energy models, GNNs, neural operators such as FNO, DeepONet, and GNO) with PDE and boundary consistency, constitutive relations, and structured sparsity or low-rank priors. The models support editable inversion, uncertainty quantification, and active re-measurement for localized statistical channel modeling and environment-aware communication.
- Neural radio radiance fields; structured Bayesian inference on graphs
- PDE and boundary consistency; material priors; calibrated uncertainty with multimodal fusion
- Geometry / materials plus RF arrays plus contextual signals
Theoretical Part
Sparse and Efficient AI CompressionFPGA
Structured sparsity, low-rank, and tensorization across data, features, models, and gradients, with unified compression and distillation for edge–cloud collaboration and federated training. HW/SW co-acceleration (CPU, GPU, FPGA) enables reliable on-device deployment in wireless applications.
- Structured sparsity and low-rank modeling
- Compression and distillation for models and gradients
- Communication-efficient training and adaptation
- Algorithm–system co-design for throughput and energy gains
Explainable AI (XAI) Theory
Principled interpretability and reliability for models: from attribution and concept-based explanations to causal and statistical guarantees, ensuring trustworthy deployment.
- Attribution and concept-based explanations; counterfactual and causal probes; manifold learning
- Generalization and convergence bounds for nonconvex / structured models
- Risk control with conformal prediction; safety constraints and robust training
Application Part
Learning-based Optimization (L2O) GNNRL
Differentiable, learning-augmented solvers (RL, GNN, and neural combinatorial optimization) operating on the WWM for constrained decisions, with feasibility guarantees, learned warm-starts, and fast cross-scenario adaptation.
- UAV sensing plans, site placement; beam, power, and spectrum allocation; routing and formation planning
- Relax–optimize–and–sample; neural projections; constraint satisfaction in deployment
- Edge and federated execution for real-time, certifiable decisions
FieldMind: From Representation to Decisions Systems
FieldMind is our instantiation of the WWM + sparse/efficient AI + L2O, coupled with algorithm–hardware co-design (C++, CUDA, FPGA; edge and federated) for online decision making. It maps structure across data, algorithms, and architectures into solvers and hardware.