Second-Order Optimization with Lazy Hessians

Authors: Nikita Doikov, El Mahdi Chayti, Martin Jaggi

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate an illustrative numerical experiment on the performance of the proposed second-order methods with lazy Hessian updates. We consider the following convex minimization problem with the Soft Maximum objective (log-sum-exp): min x Rd f(x) := ยต ln n P i=1 exp ai,x bi h ai, x bi i .
Researcher Affiliation Academia 1Machine Learning and Optimization Laboratory, EPFL, Switzerland.
Pseudocode Yes Algorithm 1 Cubic Newton with Lazy Hessians; Algorithm 2 Regularized Newton with Lazy Hessians; Algorithm 3 Adaptive Cubic Newton with Lazy Hessians
Open Source Code No The paper does not contain any statements about releasing source code or links to a code repository for the methodology described.
Open Datasets Yes Logistic regression: a9a, d = 123, n = 32561, L2-regularization; www.csie.ntu.edu.tw/ cjlin/libsvmtools/datasets/.
Dataset Splits No The paper does not explicitly provide specific training/test/validation dataset splits (e.g., exact percentages, sample counts, or citations to predefined splits) needed to reproduce the data partitioning.
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, processor types with speeds, or memory amounts) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers) needed to replicate the experiment.
Experiment Setup Yes The regularization parameter is fixed as M := 1. We also show the performance of the Gradient Method as a standard baseline. We use a constant regularization parameter M (correspondingly the stepsize in the Gradient Method), that we choose for each method separately to optimize its performance.