Comparison Between Gradient Descent and Adam Algorithms for Image Reconstruction in Diffuse Optical Tomography
摘要
In this work, we aim to solve the inverse problem of diffuse optical tomography by using enhanced gradient descent methods. The light propagation throughout the medium is described by the diffusion approximation in frequency domain. For comparison purpose we use the gradient descent method. We have studied the convergence of the objective functional. Our simulation results, in all cases we have tested, show the robustness and the quick convergence of Adam algorithm compared to the gradient descent algorithm.