Timezone: »
We approach the graph generation problem from a spectral perspective by first generating the dominant parts of the graph Laplacian spectrum and then building a graph matching these eigenvalues and eigenvectors. Spectral conditioning allows for direct modeling of the global and local graph structure and helps to overcome the expressivity and mode collapse issues of one-shot graph generators.Our novel GAN, called SPECTRE, enables the one-shot generation of much larger graphs than previously possible with one-shot models. SPECTRE outperforms state-of-the-art deep autoregressive generators in terms of modeling fidelity, while also avoiding expensive sequential generation and dependence on node ordering. A case in point, in sizable synthetic and real-world graphs SPECTRE achieves a 4-to-170 fold improvement over the best competitor that does not overfit and is 23-to-30 times faster than autoregressive generators.
Author Information
Karolis Martinkus (ETH Zurich)
Andreas Loukas (EPFL)
Nathanaël Perraudin (Swiss Data Science Center, ETH Zürich)
Roger Wattenhofer (ETH Zurich)
Related Events (a corresponding poster, oral, or spotlight)
-
2022 Spotlight: SPECTRE: Spectral Conditioning Helps to Overcome the Expressivity Limits of One-shot Graph Generators »
Wed. Jul 20th 09:55 -- 10:00 PM Room Ballroom 1 & 2
More from the Same Authors
-
2022 Poster: A Theoretical Comparison of Graph Neural Network Extensions »
Pál András Papp · Roger Wattenhofer -
2022 Spotlight: A Theoretical Comparison of Graph Neural Network Extensions »
Pál András Papp · Roger Wattenhofer -
2022 Poster: A Deep Learning Approach for the Segmentation of Electroencephalography Data in Eye Tracking Applications »
Lukas Wolf · Ard Kastrati · Martyna Plomecka · Jieming Li · Dustin Klebe · Alexander Veicht · Roger Wattenhofer · Nicolas Langer -
2022 Spotlight: A Deep Learning Approach for the Segmentation of Electroencephalography Data in Eye Tracking Applications »
Lukas Wolf · Ard Kastrati · Martyna Plomecka · Jieming Li · Dustin Klebe · Alexander Veicht · Roger Wattenhofer · Nicolas Langer -
2019 Poster: Adversarial Generation of Time-Frequency Features with application in audio synthesis »
Andrés Marafioti · Nathanaël Perraudin · Nicki Holighaus · Piotr Majdak -
2019 Oral: Adversarial Generation of Time-Frequency Features with application in audio synthesis »
Andrés Marafioti · Nathanaël Perraudin · Nicki Holighaus · Piotr Majdak -
2018 Poster: Spectrally Approximating Large Graphs with Smaller Graphs »
Andreas Loukas · Pierre Vandergheynst -
2018 Oral: Spectrally Approximating Large Graphs with Smaller Graphs »
Andreas Loukas · Pierre Vandergheynst -
2018 Poster: Fast Approximate Spectral Clustering for Dynamic Networks »
Lionel Martin · Andreas Loukas · Pierre Vandergheynst -
2018 Oral: Fast Approximate Spectral Clustering for Dynamic Networks »
Lionel Martin · Andreas Loukas · Pierre Vandergheynst -
2017 Poster: How Close Are the Eigenvectors of the Sample and Actual Covariance Matrices? »
Andreas Loukas -
2017 Talk: How Close Are the Eigenvectors of the Sample and Actual Covariance Matrices? »
Andreas Loukas