Abstract:Multidimensional scaling (MDS) is the act of embedding proximity information about a set of $n$ objects in $d$-dimensional Euclidean space. As originally conceived by the psychometric community, MDS was concerned with embedding a fixed set of proximities associated with a fixed set of objects. Modern concerns, e.g., that arise in developing asymptotic theories for statistical inference on random graphs, more typically involve studying the limiting behavior of a sequence of proximities associated with an increasing set of objects. Standard results from the theory of point-to-set maps imply that, if $n$ is fixed and a sequence of proximities converges, then the limit of the embedded structures is the embedded structure of the limiting proximities. But what if $n$ increases? It then becomes necessary to reformulate MDS so that the entire sequence of embedding problems can be viewed as a sequence of optimization problems in a fixed space. We present such a reformulation and derive some consequences.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to study the asymptotic behavior of embedding structures in Multidimensional Scaling (MDS) when the number of objects tends to infinity. Traditionally, MDS mainly focuses on the embedding problem of distance or similarity data between a fixed number of objects. However, in modern applications, such as in the fields of manifold learning and network science, the number of objects often tends to infinity, which requires new theoretical tools to handle the limit behavior of embedding structures in such cases. Specifically, the paper focuses on the following two aspects: 1. **Manifold Learning**: Study the recovery of data manifolds. Many existing theories provide asymptotic guarantees when the manifold is more densely sampled. If recovery means representation in Euclidean space, then it is natural to be concerned about the behavior of these representations in the asymptotic case. 2. **Network Science**: Study the behavior of graphs with an increasing number of vertices. By constructing a pairwise distance matrix between vertices, the problem of embedding a graph into Euclidean space can be reduced to an MDS problem. Similarly, it is natural to be concerned about the behavior of these Euclidean representations in the asymptotic case. The paper re - formulates the traditional MDS problem as a sequence of optimization problems in a fixed space by introducing the method of Continuous Multidimensional Scaling, so that it can study the limit behavior of embedding structures when the number of objects increases. This method allows the application of point - set mapping theory to analyze the asymptotic behavior and can provide theoretical guarantees even when the feasible set is infinite - dimensional. In summary, this paper aims to fill the gaps in the existing literature, especially how to systematically study and understand the asymptotic behavior of MDS embedding structures when the number of objects tends to infinity.

Continuous Multidimensional Scaling

Multidimensional scaling for big data

Multi-dimensional Scaling from K-Nearest Neighbourhood Distances

Modern Multidimensional Scaling: Theory and Applications

A dual basis approach to multidimensional scaling

A quasi-polynomial time algorithm for Multi-Dimensional Scaling via LP hierarchies

Spectral multidimensional scaling

Bayesian Hyperbolic Multidimensional Scaling

Applied Multidimensional Scaling and Unfolding

Theoretical Foundations of Ordinal Multidimensional Scaling, Including Internal and External Unfolding

A Globally Convergent Inertial First-Order Optimization Method for Multidimensional Scaling

A Supplement to Multidimensional Scaling Framework Formobile Location: a Unified View.

Progressive Glimmer: Expanding Dimensionality in Multidimensional Scaling

On eigenvalues and eigenfunctions of the operators defining multidimensional scaling on some symmetric spaces

SQuadMDS: a lean Stochastic Quartet MDS improving global structure preservation in neighbor embedding like t-SNE and UMAP

DIMAL: Deep Isometric Manifold Learning Using Sparse Geodesic Sampling

Dimension Reduction With Prior Information for Knowledge Discovery

Sparse Bayesian multidimensional scaling(s)

Multidimensional scaling of integrated neurocognitive function and schizophrenia as a disconnexion disorder.

Asymptotic behavior of the Manhattan distance in $n$-dimensions: Estimating multidimensional scenarios in empirical experiments

Understanding High Dimensional Spaces through Visual Means Employing Multidimensional Projections