Unmasking the Universe with neural nets

I will present a neural network method for reconstructing the underlying 3D cosmological density and velocity fields from discrete and incomplete observed galaxy distributions. These reconstructions are a powerful probe of cosmological parameters. One of the main aims of my talk is to demystify neural networks by clarifying their relation to different conventional statistical estimators. After discussing this, I will explicitly compare the performance of our reconstruction method with the traditional Wiener filter and highlight the advantages of the neural network approach, particularly in capturing nonlinear features. I will conclude with a discussion of the impact of neural networks on the future of the field.