期刊
2022 30TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO 2022)
卷 -, 期 -, 页码 727-731出版社
IEEE
关键词
Graph signal processing; network diffusion; deep learning; blind deconvolution; algorithm unrolling
类别
资金
- NSF [CCF-1750428, CCF-1934962, ECCS-1809356]
In this study, a deep learning solution is proposed for localizing sources of network diffusion. By leveraging graph signal processing and the ADMM method, a diffusion filter and source locations can be estimated. The trained neural network model, SLoG-Net, is interpretable, parameter efficient, and offers controllable complexity, achieving comparable performance and significant speedups compared to traditional methods.
We propose a deep learning solution to the inverse problem of localizing sources of network diffusion. Invoking graph signal processing (GSP) fundamentals, the problem boils down to blind estimation of a diffusion filter and its sparse input signal encoding the source locations. While the observations are bilinear functions of the unknowns, a mild requirement on invertibility of the graph filter enables a convex reformulation that we solve via the alternating-direction method of multipliers (ADMM). We unroll and truncate the novel ADMM iterations, to arrive at a parameterized neural network architecture for Source Localization on Graphs (SLoG-Net), that we train in an end-to-end fashion using labeled data. This way we leverage inductive biases of a GSP model-based solution in a data-driven trainable parametric architecture, which is interpretable, parameter efficient, and offers controllable complexity during inference. Experiments with simulated data corroborate that SLoG-Net exhibits performance in par with the iterative ADMM baseline, while attaining significant (post-training) speedups.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据