Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Concept Videos

Protecting Groups for Aldehydes and Ketones: Introduction01:23

Protecting Groups for Aldehydes and Ketones: Introduction

6.5K
Protecting groups are compounds that can bind to a specific functional group in the presence of other functional groups to protect them from undesired chemical reactions. These compounds can selectively bind to particular functional groups and advance chemoselective reactions in polyfunctional systems (Figure 1). After the functional group has served its purpose, it is removed by reacting it with specific compounds.
6.5K
Radical Anti-Markovnikov Addition to Alkenes: Overview01:25

Radical Anti-Markovnikov Addition to Alkenes: Overview

3.2K
The addition of hydrogen bromide to alkenes in the presence of hydroperoxides or peroxides proceeds via an anti-Markovnikov pathway and yields alkyl bromides.
3.2K
Predicting Molecular Geometry02:27

Predicting Molecular Geometry

34.0K
VSEPR Theory for Determination of Electron Pair Geometries
34.0K
Network Covalent Solids02:18

Network Covalent Solids

13.3K
Network covalent solids contain a three-dimensional network of covalently bonded atoms as found in the crystal structures of nonmetals like diamond, graphite, silicon, and some covalent compounds, such as silicon dioxide (sand) and silicon carbide (carborundum, the abrasive on sandpaper). Many minerals have networks of covalent bonds.
To break or to melt a covalent network solid, covalent bonds must be broken. Because covalent bonds are relatively strong, covalent network solids are typically...
13.3K
Radical Anti-Markovnikov Addition to Alkenes: Thermodynamics01:32

Radical Anti-Markovnikov Addition to Alkenes: Thermodynamics

1.8K
The anti-Markovnikov addition of hydrogen halides to an alkene is thermodynamically feasible only with HBr. The radical addition reaction with other hydrogen halides like HCl and HI is thermodynamically unfavorable.
1.8K
¹H NMR: Complex Splitting01:13

¹H NMR: Complex Splitting

1.2K
A proton M that is coupled to a proton X results in doublet signals for M. However, NMR-active nuclei can be simultaneously coupled to more than one nonequivalent nucleus. When M is coupled to a second proton A, such as in styrene oxide, each peak in the doublet is split into another doublet.
Splitting diagrams or splitting tree diagrams are routinely used to depict such complex couplings. While drawing splitting diagrams, the splitting with the larger coupling constant is usually applied...
1.2K

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

FreeMMIF: interactive multimodal medical image fusion via instruction-aware diffusion.

Frontiers in neurology·2026
Same author

Sirtuin 1 deficiency mediates chronic kidney disease-induced inflammaging cardiovascular calcification.

Molecular biomedicine·2026
Same author

Mechanism of palytoxin-induced ferroptosis in HaCaT cells via targeting TrxR1.

Cell biology and toxicology·2026
Same author

Mechanisms ofra Tetmethylpyrazine in spinal cord injury: a narrative review.

Molecular biology reports·2026
Same author

Self-assembled anchoring shell on NiO<sub><i>x</i></sub> nanocrystals enables efficient and stable wide-bandgap perovskite solar cells.

Chemical communications (Cambridge, England)·2026
Same author

Neural stem cell extracellular vesicle-mediated delivery of astragaloside IV attenuates hypoxic-ischemic brain damage by activating the mTOR pathway.

Naunyn-Schmiedeberg's archives of pharmacology·2026

Related Experiment Video

Updated: May 20, 2025

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

442

Mutual GNN-MLP distillation for robust graph adversarial defense.

Bowen Deng1, Jialong Chen2, Yanming Hu2

  • 1School of Computer Science and Engineering, Sun Yat-sen University, No. 132, Outer Ring East Road, Guangzhou, 510006, Guangdong, China; School of Systems Science and Engineering, Sun Yat-sen University, No. 135, Xingang West Road, Guangzhou, 510275, Guangdong, China.

Neural Networks : the Official Journal of the International Neural Network Society
|May 17, 2025
PubMed
Summary

This study introduces Mutual GNN-MLP distillation (MGMD) to improve adversarial defenses for graph neural networks (GNNs). MGMD enhances adaptability to graph heterophily and offers scalable inference, overcoming limitations of current GNN defense methods.

Keywords:
Adversarial robustnessGraph heterophilyGraph knowledge distillationGraph neural networkSemi-supervised learning

More Related Videos

Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

8.9K
Revealing Neural Circuit Topography in Multi-Color
09:11

Revealing Neural Circuit Topography in Multi-Color

Published on: November 14, 2011

14.9K

Related Experiment Videos

Last Updated: May 20, 2025

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

442
Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

8.9K
Revealing Neural Circuit Topography in Multi-Color
09:11

Revealing Neural Circuit Topography in Multi-Color

Published on: November 14, 2011

14.9K

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Graph Neural Networks

Background:

  • Current adversarial defenses for Graph Neural Networks (GNNs) exhibit limitations in adapting to graph heterophily, generalizing to early GNN models like GraphSAGE, and achieving scalable inference.
  • These limitations restrict the practical application of GNNs in resource-constrained environments and against sophisticated adversarial attacks.

Purpose of the Study:

  • To propose a novel framework, Mutual GNN-MLP distillation (MGMD), that addresses the limitations of existing GNN adversarial defenses.
  • To enhance GNN adaptability to graph heterophily and improve robustness against structure and node feature attacks.
  • To achieve high inference scalability for GNNs, making them suitable for resource-constrained scenarios.

Main Methods:

  • Developed the Mutual GNN-MLP distillation (MGMD) framework, which combines the strengths of GNNs and Multi-layer Perceptrons (MLPs).
  • Integrated GNNs and MLPs to enhance adaptability to graph heterophily and defend against adversarial attacks.
  • Introduced a novel learning rate scheduler, inspired by convergence analysis, to mitigate inductive bias conflicts between GNN and MLP components.
  • Formally demonstrated MGMD's adversarial robustness and adaptability through decision boundary analysis.

Main Results:

  • MGMD demonstrated enhanced adaptability to graph heterophily and improved adversarial robustness compared to prior methods.
  • The distilled MLP component enabled significantly high inference scalability.
  • Experiments on diverse homophilic and heterophilic graphs validated the effectiveness of the proposed learning rate scheduler.
  • MGMD showed clear advantages over existing adversarial defense methods for GNNs.

Conclusions:

  • The proposed Mutual GNN-MLP distillation (MGMD) framework effectively addresses key limitations in current GNN adversarial defenses.
  • MGMD offers a scalable and robust solution for GNNs, particularly in heterophilic graph settings and resource-constrained environments.
  • The novel learning rate scheduler contributes to stable training and improved performance of the hybrid GNN-MLP model.