{"id":"arxiv/2602.03118","url":"https://arxiv.org/abs/2602.03118","tags":["math.NA"],"score":0,"title":"The High Cost of Data Augmentation for Learning Equivariant Models","domain":"math.NA","source":"arxiv","authors":["Henri Klintebäck","Christoph Ortner","Lior Silberman"],"reviews":0,"subject":"math.NA","summary":"According to Noether's theorem the presence of a continuous symmetry in a Hamiltonian systems is equivalent to the existence of a conserved quantity, yet these symmetries are not always explicitly enforced in data-driven models. There remains a debate whether or not encoding of symmetry into a model architecture is the optimal approach. A competing approach is to target approximate symmetry through data augmentation. In this work, we study two approaches aimed at improving the symmetry properties of such an approximation scheme: one based on a quadrature rule for the Haar measure on the…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-02-03","reactions":{}}