Saved in:
Bibliographic Details
Main Authors: Wang, Ruiqi, Klabjan, Diego
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2508.01764
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912856608342016
author Wang, Ruiqi
Klabjan, Diego
author_facet Wang, Ruiqi
Klabjan, Diego
contents The concept of learning to optimize involves utilizing a trainable optimization strategy rather than relying on manually defined full gradient estimations such as ADAM. We present a framework that jointly trains the full gradient estimator and the trainable weights of the model. Specifically, we prove that pseudo-linear TO (Trainable Optimizer), a linear approximation of the full gradient, matches SGD's convergence rate while effectively reducing variance. Pseudo-linear TO incurs negligible computational overhead, requiring only minimal additional tensor multiplications. To further improve computational efficiency, we introduce two simplified variants of Pseudo-linear TO. Experiments demonstrate that TO methods converge faster than benchmark algorithms (e.g., ADAM) in both strongly convex and non-convex settings, and fine tuning of an LLM.
format Preprint
id arxiv_https___arxiv_org_abs_2508_01764
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Trainable Optimizer
Wang, Ruiqi
Klabjan, Diego
Machine Learning
The concept of learning to optimize involves utilizing a trainable optimization strategy rather than relying on manually defined full gradient estimations such as ADAM. We present a framework that jointly trains the full gradient estimator and the trainable weights of the model. Specifically, we prove that pseudo-linear TO (Trainable Optimizer), a linear approximation of the full gradient, matches SGD's convergence rate while effectively reducing variance. Pseudo-linear TO incurs negligible computational overhead, requiring only minimal additional tensor multiplications. To further improve computational efficiency, we introduce two simplified variants of Pseudo-linear TO. Experiments demonstrate that TO methods converge faster than benchmark algorithms (e.g., ADAM) in both strongly convex and non-convex settings, and fine tuning of an LLM.
title A Trainable Optimizer
topic Machine Learning
url https://arxiv.org/abs/2508.01764