Learning from Memory: Non-Parametric Memory Augmented Self-Supervised Learning of Visual Features

TL;DR

This paper presents a non-parametric memory-augmented self-supervised learning method that enhances training stability and transferability of visual features, reducing computational costs and improving performance across multiple vision tasks.

Contribution

It introduces a novel memory component and stochastic memory blocks to stabilize SSL training and improve feature transferability without extra regularizers.

Findings

Achieves stable SSL training without additional regularizers.

Produces highly transferable visual representations.

Requires less computing time and resources.

Abstract

This paper introduces a novel approach to improving the training stability of self-supervised learning (SSL) methods by leveraging a non-parametric memory of seen concepts. The proposed method involves augmenting a neural network with a memory component to stochastically compare current image views with previously encountered concepts. Additionally, we introduce stochastic memory blocks to regularize training and enforce consistency between image views. We extensively benchmark our method on many vision tasks, such as linear probing, transfer learning, low-shot classification, and image retrieval on many datasets. The experimental results consolidate the effectiveness of the proposed approach in achieving stable SSL training without additional regularizers while learning highly transferable representations and requiring less computing time and resources.