Abstract:Network embedding, aiming to embed a network into a low dimensional vector space while preserving the inherent structural properties of the network, has attracted considerable attentions recently. Most of the existing embedding methods embed nodes as point vectors in a low-dimensional continuous space. In this way, the formation of the edge is deterministic and only determined by the positions of the nodes. However, the formation and evolution of real-world networks are full of uncertainties, which makes these methods not optimal. To address the problem, we propose a novel Deep Variational Network Embedding in Wasserstein Space (DVNE) in this paper. The proposed method learns a Gaussian distribution in the Wasserstein space as the latent representation of each node, which can simultaneously preserve the network structure and model the uncertainty of nodes. Specifically, we use 2-Wasserstein distance as the similarity measure between the distributions, which can well preserve the transitivity in the network with a linear computational cost. Moreover, our method implies the mathematical relevance of mean and variance by the deep variational model, which can well capture the position of the node by the mean vectors and the uncertainties of nodes by the variance. Additionally, our method captures both the local and global network structure by preserving the first-order and second-order proximity in the network. Our experimental results demonstrate that our method can effectively model the uncertainty of nodes in networks, and show a substantial gain on real-world applications such as link prediction and multi-label classification compared with the state-of-the-art methods.

Geodesic Properties of a Generalized Wasserstein Embedding for Time Series Analysis

Integrating Data-Driven Segmentation, Local Feature Extraction and Fisher Kernel Encoding to Improve Time Series Classification

Linear Partial Gromov-Wasserstein Embedding

Geometric Characteristics of the Wasserstein Metric on SPD(n) and Its Applications on Data Processing

Learning High Dimensional Wasserstein Geodesics

Wasserstein Wormhole: Scalable Optimal Transport Distance with Transformers

Wasserstein Spatial Depth

Generalized Sliced Wasserstein Distances

Geometry and analytic properties of the sliced Wasserstein space

GeoStat Representations of Time Series for Fast Classification

Stereographic Spherical Sliced Wasserstein Distances

Information Geometry of Wasserstein Statistics on Shapes and Affine Deformations

A deep learning framework for geodesics under spherical Wasserstein-Fisher-Rao metric and its application for weighted sample generation

Methods for Space-Time Analysis and Modeling

Designing Universal Causal Deep Learning Models: The Geometric (Hyper)Transformer

Wasserstein multivariate auto-regressive models for modeling distributional time series

A Wasserstein-Type Distance for Gaussian Mixtures on Vector Bundles with Applications to Shape Analysis

Deep Variational Network Embedding in Wasserstein Space

Sliced Wasserstein Geodesics and Equivalence Wasserstein and Sliced Wasserstein metrics

Isometries and isometric embeddings of Wasserstein spaces over the Heisenberg group

A statistical framework for analyzing shape in a time series of random geometric objects