English

Control Theoretic Approach to Fine-Tuning and Transfer Learning

Machine Learning 2024-05-21 v2 Optimization and Control

Abstract

Given a training set in the form of a paired (X,Y)(\mathcal{X},\mathcal{Y}), we say that the control system x˙=f(x,u)\dot x = f(x,u) has learned the paired set via the control uu^* if the system steers each point of X\mathcal{X} to its corresponding target in Y\mathcal{Y}. If the training set is expanded, most existing methods for finding a new control uu^* require starting from scratch, resulting in a quadratic increase in complexity with the number of points. To overcome this limitation, we introduce the concept of  tuning without forgetting\textit{ tuning without forgetting}. We develop an iterative algorithm\textit{an iterative algorithm} to tune the control uu^* when the training set expands, whereby points already in the paired set are still matched, and new training samples are learned. At each update of our method, the control uu^* is projected onto the kernel of the end-point mapping generated by the controlled dynamics at the learned samples. It ensures keeping the end-points for the previously learned samples constant while iteratively learning additional samples.

Keywords

Cite

@article{arxiv.2404.11013,
  title  = {Control Theoretic Approach to Fine-Tuning and Transfer Learning},
  author = {Erkan Bayram and Shenyu Liu and Mohamed-Ali Belabbas and Tamer Başar},
  journal= {arXiv preprint arXiv:2404.11013},
  year   = {2024}
}
R2 v1 2026-06-28T15:56:38.297Z