Finding Dark Matter with Computer Vision: Probing the Subhalo Mass Function

Dark matter is over five times more abundant than all visible matter in the universe, yet nearly a century after its discovery, we still don’t know what it is. Images of gravitational lenses -- massive objects that bend light from distant galaxies -- may hold the key, encoding dark matter signatures in the distorted images. With advances in artificial intelligence, it’s natural to ask whether machine learning can help detect these hidden signatures. This thesis explores that possibility under various conditions, finding that it is a much more challenging problem than initially thought.