What question did this study set out to answer?

The aim is to enhance the efficiency of finite-sum optimization using Hessian-averaging techniques that tolerate gradient inexactness.

April 10, 2026Mathematical ProgrammingOpen Access

Fast finite-sum optimization via cyclically-sampled Hessian averaging methods

View Full Paper

Ask AI

Bookmark

Authors

TOThomas O’Leary-RoseberryThe Ohio State University RBRaghu Bollapragada

Discussion

Member takes

Implication

Demonstrates improved convergence in logistic regression via Hessian-averaging methods with gradient inexactness, highlighting practical implications.

Key Points

The aim is to enhance the efficiency of finite-sum optimization using Hessian-averaging techniques that tolerate gradient inexactness.
Considered Hessian-averaging based subsampled Newton methods for optimization.
Proposed norm condition based adaptive-sampling strategies for gradient inexactness.
Applied deterministic cyclic sampling techniques for subsampled Hessian approximations.
Developed a comprehensive convergence theory addressing both strong convexity and nonconvexity.
Achieved fast local superlinear convergence rates of O(1/k) under new methodologies.
Demonstrated significant improvements in convergence on logistic regression tasks compared to traditional stochastic sampling methods.

Cite This Study

O’Leary-Roseberry et al. (2026) studied this question.

synapsesocial.com/papers/69d895206c1944d70ce06110 https://doi.org/10.1007/s10107-026-02354-0

View Full Paper

Ask AI

Bookmark

Implication

Demonstrates improved convergence in logistic regression via Hessian-averaging methods with gradient inexactness, highlighting practical implications.

Fast finite-sum optimization via cyclically-sampled Hessian averaging methods

Discussion

Member takes

Implication

Key Points

Cite This Study

Fast finite-sum optimization via cyclically-sampled Hessian averaging methods

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider