Local transfer learning from one data space to another
This interesting paper studies transfer learning from one data space to another. The following problem is well known and fundamental in machine learning. Suppose we are given data \((x_j,y_j)\) assumed to be sampled from an unknown probability distribution. The objective is to approximate the function \(\mathcal{E}(y|x)\) from the data. Often, the points \(x_j\) belong to an ambient Euclidean space of a very high dimension, leading to the so-called curse of dimensionality. One of the strategies to counter this ``curse is to assume the manifold hypothesis that is that the points \(x_j\) live on or near to an unknown low-dimensional sub-manifold of the ambient space and there is a vast literature which covers this topic ranging from Isomaps, locally linear embeddings, Hessian locally linear embeddings, diffusion maps, semidefinite programming, local tangent space alignment, Laplacian eigenmaps, randomized anisotropic transforms and many others. The manifold is defined by the data set itself and, is designed so that the data is dense on the manifold in some well defined sense. The notion of a data space is an abstraction of a manifold encapsulating the essential properties that allow for function approximation. The problem of transfer learning is to use the learning of a function on one data set to learn a similar function on a new data set. In terms of function approximation, this means lifting a function on one data space (the base data space) to another (the target data space). In this paper, the authors study the question of such lifting when the data is assumed to be known only on a part of the base data space. The authors are interested in determining subsets of the target data space on which the lifting can be defined, and how the local smoothness of the function and its lifting are related. \N\NThe paper is well written with an excellent set of references.\N\NFor the entire collection see [Zbl 1577.65002].
- \(L^ p\)-theory of elliptic differential operators on manifolds of bounded geometry
- A generalized diffusion frame for parsimonious representation of functions on data defined manifolds
- A quadrature formula for diffusion polynomials corresponding to a generalized heat kernel
- A unified framework for harmonic analysis of functions on directed graphs and changing data
- A weighted transplantation theorem for Jacobi coefficients
- Band-limited localized Parseval frames and Besov spaces on compact homogeneous manifolds
- Community Detection Using Spectral Clustering on Sparse Geosocial Data
- Diffuse Interface Models on Graphs for Classification of High Dimensional Data
- Diffusion maps
- Diffusion maps for changing data
- Diffusion polynomial frames on metric measure spaces
- Diffusion wavelets
- Dimensionality reduction of hyperspectral imagery data for feature classification
- Discrete-continuous Jacobi-Sobolev spaces and Fourier series
- Eignets for function approximation on manifolds
- Fundamentals of radar imaging
- Gaussian upper bounds for the heat kernel on arbitrary Riemannian manifolds
- scientific article; zbMATH DE number 52737 (Why is no real title available?)
- scientific article; zbMATH DE number 3477793 (Why is no real title available?)
- scientific article; zbMATH DE number 3108589 (Why is no real title available?)
- Image manifolds which are isometric to Euclidean space
- Laplacian Eigenmaps for Dimensionality Reduction and Data Representation
- Learning Theory
- Manifold parametrizations by eigenfunctions of the Laplacian and heat kernels
- Marcinkiewicz-Zygmund measures on manifolds
- Provable approximation properties for deep neural networks
- Randomized anisotropic transform for nonlinear dimensionality reduction
- Riesz transform, Gaussian bounds and the method of wave equation
- Semi-supervised learning on Riemannian manifolds
- Synthetic aperture inversion
- The mathematics of computerized tomography
- Transferability of graph neural networks: an extended graphon approach
- Transplantation theorems and multiplier theorems for Jacobi series
- Universal local parametrizations via heat kernels and eigenfunctions of the Laplacian
- Upper bounds of derivatives of the heat kernel on an arbitrary complete manifold
- Wave equations for graphs and the edge-based Laplacian
This page was built for publication: Local transfer learning from one data space to another
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7308183)
