AN EFFICIENT ITERATIVE METHOD FOR SOLVING MULTIPLE SCATTERING IN LOCALLY INHOMOGENEOUS MEDIA USING DEEP LEARNING

Traditional methods for solving multiple scattering problems involve computationally expensive iterative techniques that require substantial computational resources and time. Our approach leverages the power of deep learning models, specifically convolutional neural networks (CNNs), to approximate the multiple scattering effects in locally inhomogeneous media. We train a CNN to learn the complex relationship between the incident waves and the scattered waves based on a dataset of simulated scattering problems. The CNN is trained to predict the scattered waves given the incident waves and the local inhomogeneities. Once the CNN is trained, it can be used as a fast and efficient forward model to solve multiple scattering problems in real-time. Given the incident waves and the local inhomogeneities, the CNN predicts the scattered waves iteratively, refining the predictions with each iteration. This iterative process allows the CNN to capture the multiple scattering effects accurately while significantly reducing the computational cost compared to traditional methods. The effectiveness of our proposed method through numerical experiments on various locally inhomogeneous media configurations. The results show that our method achieves high accuracy in predicting the scattered waves while significantly reducing the computational time compared to traditional iterative techniques. Using deep learning has the potential to revolutionize the field by enabling real-time and computationally efficient simulations in various applications, including imaging, remote sensing, and non-destructive testing