Paper Title:
Hodge-Compositional Edge Gaussian Processes
Published on:
30 October 2023
Primary Category:
Machine Learning
Paper Authors:
Maosheng Yang,
Viacheslav Borovitskiy,
Elvin Isufi
Proposes Gaussian processes for flows on network edges
Enables separate learning of gradient, curl, harmonic parts
Links processes to stochastic PDEs on network edges
Shows applications in forex, ocean currents, water networks
Edge flows on networks
This paper proposes Gaussian processes for modeling flows on the edges of networks. It introduces principled ways to define priors on edge flows that capture key properties like divergence-freeness and curl-freeness. The method enables separate learning of gradient, curl, and harmonic components.
Estimating Gaussian graphical models from partial graphs
Information flow in Gaussian processes across timescales
Gaussian Processes for High-Frequency PDE Solutions
Learning traffic patterns from road network data
Graph networks for porous media flow prediction
Learning particle interaction rules from motion data
No comments yet, be the first to start the conversation...
Sign up to comment on this paper