Flowchart illustrating our deep learning deconvolution model. The notations ℱ and ℱ⁻¹ signify the transformations from λ² to ϕ and back, respectively. Initially, samples are extracted from the observed data and used to generate a simulated dataset with a similar distribution. Next, the two datasets are combined and fed into the neural network. After the deconvolution process, the predictions F_obs^* and F_sim^* are separated once more and evaluated against their corresponding targets using the MSE loss.