Inverse problems aim to recover unknown parameters or hidden states from indirect and noisy observations, and they arise broadly in science and engineering. Recently, machine learning has emerged as a powerful framework for inverse problems, complementing and extending classical optimization- and regularization-based approaches. Focusing on electrical impedance tomography (EIT), this talk will trace recent methodological developments from physics-informed neural networks (PINNs) and neural operators to diffusion models. I will highlight the main ideas, advantages, and limitations of these approaches in addressing the challenges of EIT, and discuss remaining challenges and future directions.

---

회의 ID: 880 5040 4196