Signal Recovery

Signal recovery focuses on reconstructing signals from incomplete or noisy measurements, a crucial problem across diverse scientific fields. Current research emphasizes developing robust algorithms and model architectures, such as Bayesian methods (e.g., manifold-constrained Gaussian processes), low-rank deconvolution, and neural networks (including diffusion models and normalizing flows), to handle various challenges like sparse data, uncertain forward models, and nonlinear measurements. These advancements improve signal reconstruction accuracy and efficiency in applications ranging from image processing and speech separation to astrophysics and particle physics, impacting data analysis and interpretation across many disciplines.