Research
My research interests include machine learning and matrix recovery, multimodal localization using WiFi signals and LiDAR data, and diffusion models for image completion. I am particularly interested in combining model-based signal processing with data-driven learning to recover latent structure from incomplete or noisy observations, improve positioning through multimodal sensing, and reconstruct missing image content using diffusion-based priors.
Machine Learning & Matrix Recovery
We study how latent low-dimensional structure can be recovered reliably from limited, noisy, or partially observed data. Our work combines low-rank modeling, side information, nonconvex optimization, and statistical learning theory to improve sample efficiency and robustness.
Channel Estimation & Localization
We study multimodal localization methods that integrate wireless WiFi signals with LiDAR data. By combining complementary radio and geometric information through data-driven learning, we aim to achieve accurate and robust positioning in complex indoor and wireless environments.
Diffusion Models & Image Completion
We study diffusion models for image completion, with an emphasis on reconstructing missing or corrupted regions while preserving the known visual context. Our interests include mask-guided diffusion, text-guided inpainting, structural priors, and efficient restoration sampling.
