Exploiting data invariances is crucial for efficient learning in both artificial and biological neural circuits. Understanding how neural networks can discover appropriate represen-tations capable of harnessing the underlying symmetries of their inputs is thus crucial in machine learning and neuroscience. Convolutional neural networks, for example, were designed to exploit translation symmetry, and their capabilities triggered the first wave of deep learning successes. However, learning convolutions directly from translation -invariant data with a fully connected network has so far proven elusive. Here we show how initially fully connected neural networks solving a discrimination task can learn a convolutional structure directly from their inputs, resulting in localized, space-tiling receptive fields. These receptive fields match the filters of a convolutional network trained on the same task. By carefully designing data models for the visual scene, we show that the emergence of this pattern is triggered by the non-Gaussian, higher -order local structure of the inputs, which has long been recognized as the hallmark of natural images. We provide an analytical and numerical characterization of the pattern formation mechanism responsible for this phenomenon in a simple model and find an unexpected link between receptive field formation and tensor decomposition of higher -order input correlations. These results provide a perspective on the development of low-level feature detectors in various sensory modalities and pave the way for studying the impact of higher-order statistics on learning in neural networks.

Data-driven emergence of convolutional structure in neural networks / Ingrosso, A.; Goldt, S.. - In: PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA. - ISSN 1091-6490. - 119:40(2022), pp. 1-10. [10.1073/pnas.2201854119]

Data-driven emergence of convolutional structure in neural networks

Goldt S.
2022-01-01

Abstract

Exploiting data invariances is crucial for efficient learning in both artificial and biological neural circuits. Understanding how neural networks can discover appropriate represen-tations capable of harnessing the underlying symmetries of their inputs is thus crucial in machine learning and neuroscience. Convolutional neural networks, for example, were designed to exploit translation symmetry, and their capabilities triggered the first wave of deep learning successes. However, learning convolutions directly from translation -invariant data with a fully connected network has so far proven elusive. Here we show how initially fully connected neural networks solving a discrimination task can learn a convolutional structure directly from their inputs, resulting in localized, space-tiling receptive fields. These receptive fields match the filters of a convolutional network trained on the same task. By carefully designing data models for the visual scene, we show that the emergence of this pattern is triggered by the non-Gaussian, higher -order local structure of the inputs, which has long been recognized as the hallmark of natural images. We provide an analytical and numerical characterization of the pattern formation mechanism responsible for this phenomenon in a simple model and find an unexpected link between receptive field formation and tensor decomposition of higher -order input correlations. These results provide a perspective on the development of low-level feature detectors in various sensory modalities and pave the way for studying the impact of higher-order statistics on learning in neural networks.
2022
119
40
1
10
e2201854119
https://doi.org/10.1073/pnas.2201854119
Ingrosso, A.; Goldt, S.
File in questo prodotto:
File Dimensione Formato  
ingrosso-goldt-2022-data-driven-emergence-of-convolutional-structure-in-neural-networks.pdf

accesso aperto

Licenza: Creative commons
Dimensione 2.06 MB
Formato Adobe PDF
2.06 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.11767/135774
Citazioni
  • ???jsp.display-item.citation.pmc??? 2
  • Scopus 7
  • ???jsp.display-item.citation.isi??? 2
social impact