{"id":"arxiv/2603.05721","url":"https://arxiv.org/abs/2603.05721","tags":["math.NA"],"score":0,"title":"FlexTrace: Exchangeable Randomized Trace Estimation for Matrix Functions","domain":"math.NA","source":"arxiv","authors":["Madhusudan Madhavan","Alen Alexanderian","Arvind K. Saibaba"],"reviews":0,"subject":"math.NA","summary":"We consider the task of estimating the trace of a matrix function, ${\\rm tr}(f({\\bf A}))$, of a large symmetric positive semi-definite matrix ${\\bf A}$. This problem arises in multiple applications, including kernel methods and inverse problems. A key challenge across existing trace estimation methods is the need for matrix-vector products (matvecs) with $f({\\bf A})$, which can be very expensive. In this article, we introduce a novel trace estimator, FlexTrace, an exchangeable, single-pass method that estimates ${\\rm tr}(f({\\bf A}))$ solely using matvecs with ${\\bf A}$. We consider the case…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-03-05","reactions":{}}