Magnetism and morphology in the interstellar medium
Connor Stone (University of Montreal)
April 20, 2023
Abstract: We are entering a new era in astronomy with highly advanced telescopes planning massive observing programs. To fully take advantage of these incredible data will require both more advanced and faster methods. Recent developments in the field of Machine Learning have produced tools which offer a number of benefits beyond the domain of training Neural Networks. It is now possible to write complex analysis tools in such a way as to exploit GPU acceleration and automatic differentiation. In this seminar, I will present AutoProf, an upgraded tool for astronomical image modelling that incorporates novel techniques specifically designed for the new era of astronomy. AutoProf is capable of computing the “score,” which unlocks a variety of powerful algorithms and allows it to seamlessly interface with diffusion models in machine learning. By working together with diffusion models, we are able to solve several core ill-posed problems that have plagued astronomers for decades, see my talk to find out which ones!