Top-Tuning: a study on transfer learning for an efficient alternative to fine tuning for image classification with fast kernel methods

TL;DR

This paper investigates a transfer learning method called top-tuning, which uses pre-trained features with a fast kernel classifier, offering comparable accuracy to fine-tuning but with significantly reduced training time for image classification tasks.

Contribution

The study introduces and evaluates top-tuning as a computationally efficient alternative to fine-tuning, demonstrating its effectiveness on multiple small to medium-sized datasets.

Findings

Top-tuning achieves similar accuracy to fine-tuning.

Training time is reduced by one to two orders of magnitude.

Effective for small/medium datasets with limited resources.

Abstract

The impressive performance of deep learning architectures is associated with a massive increase in model complexity. Millions of parameters need to be tuned, with training and inference time scaling accordingly, together with energy consumption. But is massive fine-tuning always necessary? In this paper, focusing on image classification, we consider a simple transfer learning approach exploiting pre-trained convolutional features as input for a fast-to-train kernel method. We refer to this approach as \textit{top-tuning} since only the kernel classifier is trained on the target dataset. In our study, we perform more than 3000 training processes focusing on 32 small to medium-sized target datasets, a typical situation where transfer learning is necessary. We show that the top-tuning approach provides comparable accuracy with respect to fine-tuning, with a training time between one and two orders of magnitude smaller. These results suggest that top-tuning is an effective alternative to fine-tuning in small/medium datasets, being especially useful when training time efficiency and computational resources saving are crucial.