• Corpus ID: 229153615

Decimated Framelet System on Graphs and Fast G-Framelet Transforms

@article{Zheng2020DecimatedFS,
  title={Decimated Framelet System on Graphs and Fast G-Framelet Transforms},
  author={Xuebin Zheng and Bingxin Zhou and Yu Guang Wang and Xiaosheng Zhuang},
  journal={J. Mach. Learn. Res.},
  year={2020},
  volume={23},
  pages={18:1-18:68}
}
Graph representation learning has many real-world applications, from super-resolution imaging, 3D computer vision to drug repurposing, protein classification, social networks analysis. An adequate representation of graph data is vital to the learning performance of a statistical or machine learning model for graph-structured data. In this paper, we propose a novel multiscale representation system for graph data, called decimated framelets, which form a localized tight frame on the graph. The… 

How Framelets Enhance Graph Neural Networks

Compared to ReLU, shrinkage activation improves model performance on denoising and signal compression: noises in both node and structure can be significantly reduced by accurately cutting off the high-pass coefficients from framelet decomposition, and the signal can be compressed to less than half its original size with well-preserved prediction performance.

A Magnetic Framelet-Based Convolutional Neural Network for Directed Graphs

The predictive power of Framelet-MagNet, a magnetic framelet-based spectral GCNN for directed graphs (digraphs) is empirically validated over a range of state-of-the-art models in node classification, link prediction, and denoising.

Generalized energy and gradient flow via graph framelets

It is shown that a generalized energy via framelet decomposition can induce both low-frequency and high-frequency-dominated dynamics, via the separate weight matrices for different frequency components.

Spectral Transform Forms Scalable Transformer

This work learns from the philosophy of self-attention and proposes an efficient spectral-based neural unit that employs informative long-range temporal interaction and achieves state-of-the-art performance on a variety of online continuous-time dynamic graph learning tasks.

A Brief Survey of the Graph Wavelet Frame

This paper derives an MRA tight wavelet frame with three generators from the pseudosplines of type II, and analyzes that three wavelet functions have the highest possible order of vanishing moments, which matches the order of the approximationOrder of the framelet system provided by the refinable function.

A Survey on Graph Representation Learning Methods

An overview of non-GNN graph embedding methods, which are based on techniques such as random walks, temporal point processes and neural network learning methods, and GNN-based methods which are the application of deep learning on graph data are provided.

Graph Neural Network for Local Corruption Recovery

This work tackles the graph recovery problem from local poisons by a robustness representation learning by identifying regional graph perturbations and formulates a robust hidden feature representation for GNNs.

Robust Graph Representation Learning for Local Corruption Recovery

This work establishes a graph learning scheme that automati-cally detects (locally) corrupted feature attributes and recovers robust embedding for prediction tasks and can recover a robust graph representation from black-box poisoning and achieve excellent performance.

Weisfeiler and Lehman Go Topological: Message Passing Simplicial Networks

Message Passing Simplicial Networks (MPSNs), a class of models that perform message passing on simplicial complexes (SCs) are proposed, and a Simplicial Weisfeiler-Lehman (SWL) colouring procedure is introduced for distinguishing non-isomorphic SCs.

Cell graph neural networks enable the precise prediction of patient survival in gastric cancer

The study suggests the feasibility and benefits of such an artificial intelligence-powered digital staging system in diagnostic pathology and precision oncology and improves the assessment of the link between the TME spatial patterns and patient prognosis.

References

SHOWING 1-10 OF 150 REFERENCES

How Framelets Enhance Graph Neural Networks

Compared to ReLU, shrinkage activation improves model performance on denoising and signal compression: noises in both node and structure can be significantly reduced by accurately cutting off the high-pass coefficients from framelet decomposition, and the signal can be compressed to less than half its original size with well-preserved prediction performance.

Tight framelets on graphs for multiscale data analysis

The construction and applications of decimated tight framelets on graphs based on graph clustering algorithms, where a coarse-grained chain of graphs can be constructed where a suitable orthonormal eigenpair can be deduced, are discussed.

Stability of Graph Scattering Transforms

This work extends scattering transforms to network data by using multiresolution graph wavelets, whose computation can be obtained by means of graph convolutions, and proves that the resulting graph scattering transforms are stable to metric perturbations of the underlying network.

Graph Denoising with Framelet Regularizer

This paper tailors regularizers for graph data in terms of both feature and structure noises, where the objective function is efficiently solved with the alternating direction method of multipliers (ADMM).

Fast Haar Transforms for Graph Neural Networks

Spectrum-Adapted Tight Graph Wavelet and Vertex-Frequency Frames

This paper considers the problem of designing spectral graph filters for the construction of dictionaries of atoms that can be used to efficiently represent signals residing on weighted graphs, and proposes filters adapted to the distribution of graph Laplacian eigenvalues that lead to atoms with better discriminatory power.

Graph Neural Networks with Haar Transform-Based Convolution and Pooling: A Complete Guide

This work proposes a novel graph neural network, which it calls HaarNet, to predict graph labels with interrelated convolution and pooling strategies, which outperforms various existing GNN models, especially on big data sets.

Graph Spectral Image Processing

This paper overviews recent graph spectral techniques in GSP specifically for image/video processing, including image compression, image restoration, image filtering, and image segmentation.

A Unified View on Graph Neural Networks as Graph Signal Denoising

It is established mathematically that the aggregation processes in a group of representative GNN models including GCN, GAT, PPNP, and APPNP can be regarded as solving a graph denoising problem with a smoothness assumption.

Multiclass Data Segmentation Using Diffuse Interface Methods on Graphs

Two graph-based algorithms for multiclass segmentation of high-dimensional data on graphs using a graph adaptation of the classical numerical Merriman-Bence-Osher (MBO) scheme, which alternates between diffusion and thresholding are presented.
...