We propose GRAph Neural Diffusion with a source term (GRAND++) for graph deep learning with a limited number of labeled nodes, i.e., low-labeling rate. GRAND++ is a class of continuous-depth graph deep learning architectures whose theoretical underpinning is the diffusion process on graphs with a source term. The source term guarantees two interesting theoretical properties of GRAND++: (i) the representation of graph nodes, under the dynamics of GRAND++, will not converge to a constant vector over all nodes even as the time goes to infinity, which mitigates the over-smoothing issue of graph neural networks and enables graph learning in very deep architectures. (ii) GRAND++ can provide accurate classification even when the model is trained with a very limited number of labeled training data. We experimentally verify the above two advantages on various graph deep learning benchmark tasks, showing a significant improvement over many existing graph neural networks.
more »
« less
This content will become publicly available on March 25, 2025
Multiple-Source Localization from a Single-Snapshot Observation Using Graph Bayesian Optimization
Due to the significance of its various applications, source localization has garnered considerable attention as one of the most important means to confront diffusion hazards. Multi-source localization from a single-snapshot observation is especially relevant due to its prevalence. However, the inherent complexities of this problem, such as limited information, interactions among sources, and dependence on diffusion models, pose challenges to resolution. Current methods typically utilize heuristics and greedy selection, and they are usually bonded with one diffusion model. Consequently, their effectiveness is constrained.To address these limitations, we propose a simulation-based method termed BOSouL. Bayesian optimization (BO) is adopted to approximate the results for its sample efficiency. A surrogate function models uncertainty from the limited information. It takes sets of nodes as the input instead of individual nodes. BOSouL can incorporate any diffusion model in the data acquisition process through simulations. Empirical studies demonstrate that its performance is robust across graph structures and diffusion models. The code is available at https://github.com/XGraph-Team/BOSouL.
more »
« less
- Award ID(s):
- 2153369
- PAR ID:
- 10505922
- Publisher / Repository:
- Association for the Advancement of Artificial Intelligence
- Date Published:
- Journal Name:
- Proceedings of the AAAI Conference on Artificial Intelligence
- Volume:
- 38
- Issue:
- 20
- ISSN:
- 2159-5399
- Page Range / eLocation ID:
- 22538 to 22546
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
We propose GRAph Neural Diffusion with a source term (GRAND++) for graph deep learning with a limited number of labeled nodes, i.e., low-labeling rate. GRAND++ is a class of continuous-depth graph deep learning architectures whose theoretical underpinning is the diffusion process on graphs with a source term. The source term guarantees two interesting theoretical properties of GRAND++: (i) the representation of graph nodes, under the dynamics of GRAND++, will not converge to a constant vector over all nodes even as the time goes to infinity, which mitigates the over-smoothing issue of graph neural networks and enables graph learning in very deep architectures. (ii) GRAND++ can provide accurate classification even when the model is trained with a very limited number of labeled training data. We experimentally verify the above two advantages on various graph deep learning benchmark tasks, showing a significant improvement over many existing graph neural networks.more » « less
-
null (Ed.)Cyber-physical-social systems (CPSS) are physical devices with highly integrated functions of sensing, computing, communication and control, and are seamlessly embedded in human society. The levels of intelligence and functions that CPSS can perform rely on their extensive collaboration and information sharing through networks. In this paper, information diffusion within CPSS networks is studied. Information dynamics models are proposed to characterize the evolution of information processing and decision making capabilities of individual CPSS nodes. The data-driven statistical models are based on a mesoscale probabilistic graph model, where the individual nodes' sensing and computing functions are represented as the probabilities of correct predictions, whereas the communication functions are represented as the probabilities of mutual influences between nodes. A copula dynamics model is proposed to explicitly capture the information dependency among individuals with joint prediction probabilities and estimated from extremal probabilities. A topology-informed vector autoregression model is proposed to represent the mutual influence of prediction capabilities. A spatial-temporal hybrid Gaussian process regression model is also proposed to simultaneously capture correlations between nodes and temporal correlation in the time series.more » « less
-
Finite element analysis is used to study brain axonal injury and develop Brain White Matter (BWM) models while accounting for both the strain magnitude and the strain rate. These models are becoming more sophisticated and complicated due to the complex nature of the BMW composite structure with different material properties for each constituent phase. State-of-the-art studies focus on employing techniques that combine information about the local axonal directionality in different areas of the brain with diagnostic tools such as Diffusion-Weighted Magnetic Resonance Imaging (Diffusion-MRI). The diffusion-MRI data offers localization and orientation information of axonal tracks which are analyzed in finite element models to simulate virtual loading scenarios. Here, a BMW biphasic material model comprised of axons and neuroglia is considered. The model’s architectural anisotropy represented by a multitude of axonal orientations, that depend on specific brain regions, adds to its complexity. During this effort, we develop a finite element method to merge micro-scale Representative Volume Elements (RVEs) with orthotropic frequency domain viscoelasticity to an integrated macro-scale BWM finite element model, which incorporates local axonal orientation. Previous studies of this group focused on building RVEs that combined different volume fractions of axons and neuroglia and simulating their anisotropic viscoelastic properties. Via the proposed model, we can assign material properties and local architecture on each element based on the information from the orientation of the axonal traces. Consecutively, a BWM finite element model is derived with fully defined both material properties and material orientation. The frequency-domain dynamic response of the BMW model is analyzed to simulate larger scale diagnostic modalities such as MRI and MRE.more » « less
-
Inverse molecular generation is an essential task for drug discovery, and generative models offer a very promising avenue, especially when diffusion models are used. Despite their great success, existing methods are inherently limited by the lack of a semantic latent space that can not be navigated and perform targeted exploration to generate molecules with desired properties. Here, we present a property-guided diffusion model for generating desired molecules, which incorporates a sophisticated diffusion process capturing intricate interactions of nodes and edges within molecular graphs and leverages a time-dependent molecular property classifier to integrate desired properties into the diffusion sampling process. Furthermore, we extend our model to a multi-property-guided paradigm. Experimental results underscore the competitiveness of our approach in molecular generation, highlighting its superiority in generating desired molecules without the need for additional optimization steps.more » « less
