Unsupervised One-shot Learning of Both Specific Instances and Generalised Classes with a Hippocampal Architecture

TL;DR

This paper introduces an unsupervised neural network model inspired by hippocampal architecture that can learn both specific instances and generalized classes from a single exposure, outperforming baselines under noisy conditions.

Contribution

It extends the Omniglot benchmark to test both classification and instance recognition, implementing an unsupervised CLS-based model that rivals supervised methods without domain-specific biases.

Findings

Achieves comparable classification accuracy to supervised models on Omniglot

Significantly outperforms nearest-neighbor baseline in instance recognition with noise and occlusion

Demonstrates the ability to learn and recall specific instances in an unsupervised manner

Abstract

Established experimental procedures for one-shot machine learning do not test the ability to learn or remember specific instances of classes, a key feature of animal intelligence. Distinguishing specific instances is necessary for many real-world tasks, such as remembering which cup belongs to you. Generalisation within classes conflicts with the ability to separate instances of classes, making it difficult to achieve both capabilities within a single architecture. We propose an extension to the standard Omniglot classification-generalisation framework that additionally tests the ability to distinguish specific instances after one exposure and introduces noise and occlusion corruption. Learning is defined as an ability to classify as well as recall training samples. Complementary Learning Systems (CLS) is a popular model of mammalian brain regions believed to play a crucial role in learning from a single exposure to a stimulus. We created an artificial neural network implementation of CLS and applied it to the extended Omniglot benchmark. Our unsupervised model demonstrates comparable performance to existing supervised ANNs on the Omniglot classification task (requiring generalisation), without the need for domain-specific inductive biases. On the extended Omniglot instance-recognition task, the same model also demonstrates significantly better performance than a baseline nearest-neighbour approach, given partial occlusion and noise.