arXiv cs.AI / cs.LG / cs.CL·10d agoDouble descent is the principle of least action#double-descent#generalization#learning-theoryAI research