Deep Features for CBIR with Scarce Data using Hebbian Learning

TL;DR

This paper investigates the use of biologically inspired Hebbian learning algorithms for developing feature extractors in Content Based Image Retrieval, especially effective in scenarios with scarce labeled data, by combining unsupervised pre-training and supervised fine-tuning.

Contribution

It introduces a semi-supervised learning approach using Hebbian pre-training and SGD fine-tuning for CBIR, demonstrating improved performance with limited labeled data.

Findings

Hebbian pre-training enhances feature extraction in CBIR.

The approach outperforms alternatives with scarce labeled data.

Effective on CIFAR10 and CIFAR100 datasets.

Abstract

Features extracted from Deep Neural Networks (DNNs) have proven to be very effective in the context of Content Based Image Retrieval (CBIR). In recent work, biologically inspired \textit{Hebbian} learning algorithms have shown promises for DNN training. In this contribution, we study the performance of such algorithms in the development of feature extractors for CBIR tasks. Specifically, we consider a semi-supervised learning strategy in two steps: first, an unsupervised pre-training stage is performed using Hebbian learning on the image dataset; second, the network is fine-tuned using supervised Stochastic Gradient Descent (SGD) training. For the unsupervised pre-training stage, we explore the nonlinear Hebbian Principal Component Analysis (HPCA) learning rule. For the supervised fine-tuning stage, we assume sample efficiency scenarios, in which the amount of labeled samples is just a small fraction of the whole dataset. Our experimental analysis, conducted on the CIFAR10 and CIFAR100 datasets shows that, when few labeled samples are available, our Hebbian approach provides relevant improvements compared to various alternative methods.