Self-assignment flows for unsupervised data labeling on graphs
From MaRDI portal
assignment manifolddynamical systemsevolutionary game dynamicsgraph partitioningimage labelinginformation geometryreplicator equationspatially regularized clusteringunsupervised learning
Classification and discrimination; cluster analysis (statistical aspects) (62H30) Image analysis in multivariate analysis (62H35) Neural nets and related approaches to inference from stochastic processes (62M45) Learning and adaptive systems in artificial intelligence (68T05) Evolutionary games (91A22)
Abstract: This paper extends the recently introduced assignment flow approach for supervised image labeling to unsupervised scenarios where no labels are given. The resulting self-assignment flow takes a pairwise data affinity matrix as input data and maximizes the correlation with a low-rank matrix that is parametrized by the variables of the assignment flow, which entails an assignment of the data to themselves through the formation of latent labels (feature prototypes). A single user parameter, the neighborhood size for the geometric regularization of assignments, drives the entire process. By smooth geodesic interpolation between different normalizations of self-assignment matrices on the positive definite matrix manifold, a one-parameter family of self-assignment flows is defined. Accordingly, our approach can be characterized from different viewpoints, e.g. as performing spatially regularized, rank-constrained discrete optimal transport, or as computing spatially regularized normalized spectral cuts. Regarding combinatorial optimization, our approach successfully determines completely positive factorizations of self-assignments in large-scale scenarios, subject to spatial regularization. Various experiments including the unsupervised learning of patch dictionaries using a locally invariant distance function, illustrate the properties of the approach.
Recommendations
Cites work
- A projection technique for partitioning the nodes of a graph
- Building a completely positive factorization
- Community structure in social and biological networks
- Completely positive matrices: real, rational, and integral
- Computational optimal transport. With applications to data sciences
- Evolutionary game dynamics
- Fast partitioning of vector-valued images
- Generalized Schur complements
- Geometric approximation algorithms
- Handbook of Variational Methods for Nonlinear Geometric Data
- scientific article; zbMATH DE number 6118218 (Why is no real title available?)
- scientific article; zbMATH DE number 3862476 (Why is no real title available?)
- scientific article; zbMATH DE number 5131267 (Why is no real title available?)
- scientific article; zbMATH DE number 44578 (Why is no real title available?)
- scientific article; zbMATH DE number 734901 (Why is no real title available?)
- scientific article; zbMATH DE number 1560711 (Why is no real title available?)
- scientific article; zbMATH DE number 5223994 (Why is no real title available?)
- Image labeling by assignment
- Kernel methods in machine learning
- Low-rank doubly stochastic matrix decomposition for cluster analysis
- On the Nyström method for approximating a gram matrix for improved kernel-based learning
- On the Riemannian geometry defined by self-concordant barriers and interior-point methods.
- Optimal Transport
- Optimal transport for applied mathematicians. Calculus of variations, PDEs, and modeling
- Optimization and dynamical systems
- Revisiting the Nyström method for improved large-scale machine learning
- Riemannian geometry and geometric analysis
- SymNMF: nonnegative low-rank approximation of a similarity matrix for graph clustering
- Unsupervised assignment flow: label learning on feature manifolds by spatially regularized geometric assignment
Cited in
(8)- Assignment flow for order-constrained OCT segmentation
- Assignment flows for data labeling on graphs: convergence and stability
- Iterative multiplicative filters for data labeling
- Unsupervised assignment flow: label learning on feature manifolds by spatially regularized geometric assignment
- Assignment flows
- scientific article; zbMATH DE number 1928652 (Why is no real title available?)
- A Nonlocal Graph-PDE and Higher-Order Geometric Integration for Image Labeling
- Unsupervised labeling by geometric and spatially regularized self-assignment
This page was built for publication: Self-assignment flows for unsupervised data labeling on graphs
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5143287)
