REVIEW 2 cited by
Hutch++: Optimal Stochastic Trace Estimation
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Hutch++: Optimal Stochastic Trace Estimation
read the original abstract
We study the problem of estimating the trace of a matrix $A$ that can only be accessed through matrix-vector multiplication. We introduce a new randomized algorithm, Hutch++, which computes a $(1 \pm \epsilon)$ approximation to $tr(A)$ for any positive semidefinite (PSD) $A$ using just $O(1/\epsilon)$ matrix-vector products. This improves on the ubiquitous Hutchinson's estimator, which requires $O(1/\epsilon^2)$ matrix-vector products. Our approach is based on a simple technique for reducing the variance of Hutchinson's estimator using a low-rank approximation step, and is easy to implement and analyze. Moreover, we prove that, up to a logarithmic factor, the complexity of Hutch++ is optimal amongst all matrix-vector query algorithms, even when queries can be chosen adaptively. We show that it significantly outperforms Hutchinson's method in experiments. While our theory mainly requires $A$ to be positive semidefinite, we provide generalized guarantees for general square matrices, and show empirical gains in such applications.
Forward citations
Cited by 2 Pith papers
-
Counting Triangles of Graphs via Randomized Trace Estimation with Incomplete Matrix-Vector Products
A new partial-observation Hutchinson trace estimator for unbiased triangle counting in graphs, with variance bounds, sample complexity, and experiments showing reduced synchronization costs.
-
Towards Efficient Instanton Rate Calculations using Machine Learning Surrogates
A GPR-accelerated line integral string method makes instanton-path force evaluations nearly independent of bead count and reduces Hessian cost via selective flexible/rigid mode training.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.