Abstract:A temporal network -- a collection of snapshots recording the evolution of a network whose links appear and disappear dynamically -- can be interpreted as a trajectory in graph space. In order to characterize the complex dynamics of such trajectory via the tools of time series analysis and signal processing, it is sensible to preprocess the trajectory by embedding it in a low-dimensional Euclidean space. Here we argue that, rather than the topological structure of each network snapshot, the main property of the trajectory that needs to be preserved in the embedding is the relative graph distance between snapshots. This idea naturally leads to dimensionality reduction approaches that explicitly consider relative distances, such as Multidimensional Scaling (MDS) or identifying the distance matrix as a feature matrix in which to perform Principal Component Analysis (PCA). This paper provides a comprehensible methodology that illustrates this approach. Its application to a suite of generative network trajectory models and empirical data certify that nontrivial dynamical properties of the network trajectories are preserved already in their scalar embeddings, what enables the possibility of performing time series analysis in temporal networks.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to embed the trajectories of temporal networks into a low - dimensional Euclidean space, so as to characterize the complex dynamics of these trajectories through time - series analysis and signal - processing tools. Specifically, the paper focuses on how to preserve the relative graph distances between network snapshots during the embedding process, thereby enabling effective time - series analysis in the low - dimensional embedding. ### Core of the Problem 1. **Complex Dynamics of Temporal Networks**: A temporal network refers to the dynamic appearance and disappearance of links in the network over time. Such a network can be interpreted as a trajectory in the graph space. 2. **Selection of Embedding Methods**: In order to use time - series analysis and signal - processing tools to study these trajectories, it is necessary to embed the high - dimensional network trajectories into a low - dimensional space. The key lies in how to select the embedding method to ensure that the low - dimensional representation after embedding can preserve the key statistical properties of the original trajectory. 3. **Preserving Relative Graph Distances**: The paper emphasizes that during the embedding process, the most important thing is not to preserve the topological structure of each network snapshot, but to preserve the relative graph distances between snapshots. This can be achieved through methods such as multidimensional scaling (MDS) or principal component analysis (PCA). ### Solutions The paper proposes several methods based on linear dimensionality - reduction techniques to obtain low - dimensional (especially scalar) embeddings of network trajectories: 1. **PCA - based strategies**: - **PCA - projection strategy**: Project the feature vectors of each network snapshot onto the first few principal components to obtain a low - dimensional embedding. - **PCA - embedding strategy**: Directly use the results of principal component decomposition and take the scaled principal components as embedding coordinates. 2. **MDS - based strategies**: - **Classical - MDS strategy**: Reconstruct the hidden inner - product space through double - centering transformation, and then perform spectral decomposition to obtain a low - dimensional embedding. - **Metric - MDS strategy**: Minimize the mismatch between the pairwise distances in the embedding space and the original distance matrix through an optimization method. ### Verification Methods To verify the effectiveness of these methods, the paper uses three tools: 1. **Correlation Coefficients**: For one - dimensional time - series, calculate the Pearson correlation coefficient \( r \) and the Spearman correlation coefficient \( \rho \) between the embedding results and the original signal. 2. **Autocorrelation Function**: For network trajectories, compare the network autocorrelation function \( nACF(\tau) \) of the original network trajectory with the autocorrelation function \( ACF(\tau) \) of the scalar time - series after embedding. 3. **Maximum Lyapunov Exponent**: For models that exhibit sensitive dependence on initial conditions, compare the maximum Lyapunov exponent \( nMLE \) of the original network trajectory with the maximum Lyapunov exponent \( MLE \) of the scalar time - series after embedding. ### Conclusions The paper shows the application effects of these methods on synthetic network trajectories with different complexities and real - data, proving that low - dimensional embeddings can well preserve the intrinsic dynamic characteristics of network trajectories, thereby providing effective tools for time - series analysis of temporal networks. Through these methods, researchers can conduct more in - depth analysis of temporal networks in a low - dimensional space and reveal their complex dynamic behaviors.

Scalar embedding of temporal network trajectories

An embedding-based distance for temporal graphs

Learning dynamic embeddings for temporal attributed networks

Spatio-Temporal Trajectory Similarity Learning in Road Networks

Spatial-temporal fusion graph framework for trajectory similarity computation

MTS2Graph: Interpretable Multivariate Time Series Classification with Temporal Evolving Graphs

Statistical analysis of trajectories on Riemannian manifolds: Bird migration, hurricane tracking and video surveillance

Encoding Agent Trajectories as Representations with Sequence Transformers

Multiscale dynamical embeddings of complex networks

Another Vertical View: A Hierarchical Network for Heterogeneous Trajectory Prediction via Spectrums

Discrete-time Temporal Network Embedding via Implicit Hierarchical Learning in Hyperbolic Space

$\text{H}^2\text{TNE}$: Temporal Heterogeneous Information Network Embedding in Hyperbolic Spaces

Graph-Level Embedding for Time-Evolving Graphs

Unsupervised embedding of trajectories captures the latent structure of scientific migration

Multilayer Quantile Graph for Multivariate Time Series Analysis and Dimensionality Reduction

Dynamic Graph Embedding via LSTM History Tracking

TimeLighting: Guided Exploration of 2D Temporal Network Projections

Developing a Spatial-Temporal Contextual and Semantic Trajectory Clustering Framework

Spatio-Temporal Graph Convolutional Networks via View Fusion for Trajectory Data Analytics

Dynamic Chain Graph Models for Ordinal Time Series Data

Representation and analysis of time-series data via deep embedding and visual exploration