Convolutional Neural Networks Can (Meta-)Learn the Same-Different Relation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Gupta, Max, Rane, Sunayana, McCoy, R. Thomas, Griffiths, Thomas L.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914026854809600
author Gupta, Max
Rane, Sunayana
McCoy, R. Thomas
Griffiths, Thomas L.
author_facet Gupta, Max
Rane, Sunayana
McCoy, R. Thomas
Griffiths, Thomas L.
contents While convolutional neural networks (CNNs) have come to match and exceed human performance in many settings, the tasks these models optimize for are largely constrained to the level of individual objects, such as classification and captioning. Humans remain vastly superior to CNNs in visual tasks involving relations, including the ability to identify two objects as `same' or `different'. A number of studies have shown that while CNNs can be coaxed into learning the same-different relation in some settings, they tend to generalize poorly to other instances of this relation. In this work we show that the same CNN architectures that fail to generalize the same-different relation with conventional training are able to succeed when trained via meta-learning, which explicitly encourages abstraction and generalization across tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2503_23212
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Convolutional Neural Networks Can (Meta-)Learn the Same-Different Relation
Gupta, Max
Rane, Sunayana
McCoy, R. Thomas
Griffiths, Thomas L.
Computer Vision and Pattern Recognition
Machine Learning
68T07
I.2.0; I.2.6
While convolutional neural networks (CNNs) have come to match and exceed human performance in many settings, the tasks these models optimize for are largely constrained to the level of individual objects, such as classification and captioning. Humans remain vastly superior to CNNs in visual tasks involving relations, including the ability to identify two objects as `same' or `different'. A number of studies have shown that while CNNs can be coaxed into learning the same-different relation in some settings, they tend to generalize poorly to other instances of this relation. In this work we show that the same CNN architectures that fail to generalize the same-different relation with conventional training are able to succeed when trained via meta-learning, which explicitly encourages abstraction and generalization across tasks.
title Convolutional Neural Networks Can (Meta-)Learn the Same-Different Relation
topic Computer Vision and Pattern Recognition
Machine Learning
68T07
I.2.0; I.2.6
url https://arxiv.org/abs/2503.23212