{"id":"arxiv/2601.10737","url":"https://arxiv.org/abs/2601.10737","tags":["eess.SP","cond-mat.mtrl-sci","cs.CV","math.OC"],"score":0,"title":"Differentiating through binarized topology changes: Second-order subpixel-smoothed projection","domain":"eess.SP","source":"arxiv","authors":["Giuseppe Romano","Rodrigo Arrieta","Steven G. Johnson"],"reviews":0,"subject":"eess.SP","summary":"A key challenge in topology optimization (TopOpt) is that manufacturable structures, being inherently binary, are non-differentiable, creating a fundamental tension with gradient-based optimization. The subpixel-smoothed projection (SSP) method addresses this issue by smoothing sharp interfaces at the subpixel level through a first-order expansion of the filtered field. However, SSP does not guarantee differentiability under topology changes, such as the merging of two interfaces, and therefore violates the convergence guarantees of many popular gradient-based optimization algorithms. We…","category":"research","comments":0,"disputed":false,"division":"eess","mentions":0,"discussed":0,"published":"2026-01-08","reactions":{}}