{"id":"arxiv/2610.01050","url":"https://arxiv.org/abs/2610.01050","tags":["math.ST","stat.ME","stat.ML"],"score":0,"title":"Gradient-Guided Density Peak Clustering","domain":"math.ST","source":"arxiv","authors":["Yikun Zhang","Yen-Chi Chen"],"reviews":0,"subject":"math.ST","summary":"Density peak clustering (DPC) connects each observation to its nearest neighbor of higher density and identifies cluster centers as high-density observations with unusually large nearest neighbor uphill shifts. The resulting uphill paths from observations to cluster centers, however, can be irregular and unstable in low-density regions, making the clustering assignments sensitive to local perturbations and obscuring the population geometry of the DPC graph. In this paper, we introduce \\emph{gradient-guided density peak clustering} (GGDPC), which performs a gradient ascent step before each…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-01","reactions":{}}