Non-Negative Connectivity Causes Bow-Tie Architecture in Neural Circuits

Zhaofan Liu; CongCong Du; KongFatt Wong-Lin; Da-Hui Wang

doi:10.3389/fncir.2025.1574877

Non-Negative Connectivity Causes Bow-Tie Architecture in Neural Circuits

Zhaofan Liu
, CongCong Du
, KongFatt Wong-Lin
, Da-Hui Wang

Research output: Contribution to journal › Article › peer-review

20 Downloads (Pure)

Abstract

Bow-tie architecture (BTA) is widely observed in biological neural systems, yet the underlying mechanism driving its spontaneous emergence remains unclear. In this study, we identify a novel formation mechanism by training multi-layer neural networks under biologically inspired non-negative connectivity constraints across diverse classification tasks. We show that non-negative weights reshape network dynamics by amplifying back-propagated error signals and suppressing hidden-layer activity, leading to the self-organization of BTA without pre-defined architecture. To our knowledge, this is the first demonstration that non-negativity alone can induce BTA formation. The resulting architecture confers distinct functional advantages, including lower wiring cost, robustness to scaling, and task generalizability, highlighting both its computational efficiency and biological relevance. Our findings offer a mechanistic account of BTA emergence and bridge biological structure with
artificial learning principles.

Original language	English
Article number	1574877
Pages (from-to)	1-13
Number of pages	13
Journal	Frontiers in Neural Circuits
Volume	19
Early online date	18 Aug 2025
DOIs	https://doi.org/10.3389/fncir.2025.1574877
Publication status	Published online - 18 Aug 2025

Bibliographical note

Publisher Copyright:
Copyright © 2025 Liu, Du, Wong-Lin and Wang.

Data Availability Statement

Source code and data available at:
https://github.com/lzf531/Bow-tie-structure

Funding

The author(s) declare that financial support was received for the research and/or publication of this article. This work was funded by National Science Foundation of China (NSFC) under grant 32171094 (D-HW). KW-L was supported by HSC R&D (STL/5540/19) and MRC (MC_OC_20020).

Funders	Funder number
National Natural Science Foundation of China	32171094
HSC R&D	STL/5540/19
Medical Research Council	MC_OC_20020

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being
SDG 9 Industry, Innovation, and Infrastructure
SDG 12 Responsible Consumption and Production

Keywords

Bow-tie neural architecture
decision making
sensory classification
emergent structure
neural network model
Self-organization
learning
efficiency
bow-tie architecture
backpropagation algorithm
neural circuits
robustness
computational neuroscience
discrimination tasks
non-negative connectivity
Neural Networks, Computer
Humans
Animals
Models, Neurological
Nerve Net/physiology
Backpropagation Algorithm
Bow-tie Architecture
Efficiency
Robustness
Discrimination Tasks
Non-Negative Connectivity
Computational Neuroscience
Neural Circuits
Nerve Net

Access to Document

10.3389/fncir.2025.1574877

fncir-1-1574877 (1)Final published version, 3.92 MBLicence: CC BY

Cite this

@article{99aaa6ab20044d368aea47a43043086f,

title = "Non-Negative Connectivity Causes Bow-Tie Architecture in Neural Circuits",

abstract = "Bow-tie architecture (BTA) is widely observed in biological neural systems, yet the underlying mechanism driving its spontaneous emergence remains unclear. In this study, we identify a novel formation mechanism by training multi-layer neural networks under biologically inspired non-negative connectivity constraints across diverse classification tasks. We show that non-negative weights reshape network dynamics by amplifying back-propagated error signals and suppressing hidden-layer activity, leading to the self-organization of BTA without pre-defined architecture. To our knowledge, this is the first demonstration that non-negativity alone can induce BTA formation. The resulting architecture confers distinct functional advantages, including lower wiring cost, robustness to scaling, and task generalizability, highlighting both its computational efficiency and biological relevance. Our findings offer a mechanistic account of BTA emergence and bridge biological structure with artificial learning principles. ",

keywords = "Bow-tie neural architecture, decision making, sensory classification, emergent structure, neural network model, Self-organization, learning, efficiency, bow-tie architecture, backpropagation algorithm, neural circuits, robustness, computational neuroscience, discrimination tasks, non-negative connectivity, Neural Networks, Computer, Humans, Animals, Models, Neurological, Nerve Net/physiology, Backpropagation Algorithm, Bow-tie Architecture, Efficiency, Robustness, Discrimination Tasks, Non-Negative Connectivity, Computational Neuroscience, Neural Circuits, Nerve Net",

author = "Zhaofan Liu and CongCong Du and KongFatt Wong-Lin and Da-Hui Wang",

note = "Publisher Copyright: Copyright {\textcopyright} 2025 Liu, Du, Wong-Lin and Wang.",

year = "2025",

month = aug,

day = "18",

doi = "10.3389/fncir.2025.1574877",

language = "English",

volume = "19",

pages = "1--13",

journal = "Frontiers in Neural Circuits",

issn = "1662-5110",

publisher = "Frontiers",

}

TY - JOUR

T1 - Non-Negative Connectivity Causes Bow-Tie Architecture in Neural Circuits

AU - Liu, Zhaofan

AU - Du, CongCong

AU - Wong-Lin, KongFatt

AU - Wang, Da-Hui

PY - 2025/8/18

Y1 - 2025/8/18

N2 - Bow-tie architecture (BTA) is widely observed in biological neural systems, yet the underlying mechanism driving its spontaneous emergence remains unclear. In this study, we identify a novel formation mechanism by training multi-layer neural networks under biologically inspired non-negative connectivity constraints across diverse classification tasks. We show that non-negative weights reshape network dynamics by amplifying back-propagated error signals and suppressing hidden-layer activity, leading to the self-organization of BTA without pre-defined architecture. To our knowledge, this is the first demonstration that non-negativity alone can induce BTA formation. The resulting architecture confers distinct functional advantages, including lower wiring cost, robustness to scaling, and task generalizability, highlighting both its computational efficiency and biological relevance. Our findings offer a mechanistic account of BTA emergence and bridge biological structure with artificial learning principles.

AB - Bow-tie architecture (BTA) is widely observed in biological neural systems, yet the underlying mechanism driving its spontaneous emergence remains unclear. In this study, we identify a novel formation mechanism by training multi-layer neural networks under biologically inspired non-negative connectivity constraints across diverse classification tasks. We show that non-negative weights reshape network dynamics by amplifying back-propagated error signals and suppressing hidden-layer activity, leading to the self-organization of BTA without pre-defined architecture. To our knowledge, this is the first demonstration that non-negativity alone can induce BTA formation. The resulting architecture confers distinct functional advantages, including lower wiring cost, robustness to scaling, and task generalizability, highlighting both its computational efficiency and biological relevance. Our findings offer a mechanistic account of BTA emergence and bridge biological structure with artificial learning principles.

KW - Bow-tie neural architecture

KW - decision making

KW - sensory classification

KW - emergent structure

KW - neural network model

KW - Self-organization

KW - learning

KW - efficiency

KW - bow-tie architecture

KW - backpropagation algorithm

KW - neural circuits

KW - robustness

KW - computational neuroscience

KW - discrimination tasks

KW - non-negative connectivity

KW - Neural Networks, Computer

KW - Humans

KW - Animals

KW - Models, Neurological

KW - Nerve Net/physiology

KW - Backpropagation Algorithm

KW - Bow-tie Architecture

KW - Efficiency

KW - Robustness

KW - Discrimination Tasks

KW - Non-Negative Connectivity

KW - Computational Neuroscience

KW - Neural Circuits

KW - Nerve Net

UR - https://www.frontiersin.org/journals/neural-circuits/articles/10.3389/fncir.2025.1574877/abstract

UR - https://www.scopus.com/pages/publications/105014718900

U2 - 10.3389/fncir.2025.1574877

DO - 10.3389/fncir.2025.1574877

M3 - Article

C2 - 40900765

SN - 1662-5110

VL - 19

SP - 1

EP - 13

JO - Frontiers in Neural Circuits

JF - Frontiers in Neural Circuits

M1 - 1574877

ER -

Non-Negative Connectivity Causes Bow-Tie Architecture in Neural Circuits

Abstract

Bibliographical note

Data Availability Statement

Funding

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this