Convergence analysis of online algorithms for vector-valued kernel regression
This interesting paper presents a detailed convergence analysis of various online algorithms for vector-valued kernel regression. The authors are interested in learning regression functions from noisy vector-valued data using a certain Reproducing Hilbert Space as a prior. The paper provides a comprehensive reference list for background on the theory of kernel methods in particular the vector-valued case and we refer the interested reader to this in the paper. We now make this more mathematically formal. The authors are interested in approximating the regression function say \(g_{\nu}:K\to X\) from noisy \(\nu\)-distributed vector-valued data \((k_m, x_m)\in K\times X\) by an online learning algorithm using a reproducing kernel Hilbert space \(H\) as a prior. Assuming that the regression function essentially belongs to \(H\), the authors establish the following order-optimal estimate: \({\mathcal{E}}\left(||\varepsilon^{(m)}||_H^2\right) \leq C(m+1)^{\frac{-s}{2+s}}\), \(m=1,2,\ldots\) where \(\varepsilon^{(m)}\) denotes the error term after \(m\) processed data, the parameter \(0<s\leq 1\) expresses an additional smoothness assumption on the regression function, and the constant \(C\) depends on the variance of the input noise, the smoothness of the regression function, and other parameters of the algorithm.\N\NThe paper is well written with a good set of references.
- An introduction to the theory of reproducing kernel Hilbert spaces
- Distributed learning with regularized least squares
- Fast and strong convergence of online learning algorithms
- Harder, Better, Faster, Stronger Convergence Rates for Least-Squares Regression
- scientific article; zbMATH DE number 1332320 (Why is no real title available?)
- Kernels for vector-valued functions: a review
- Learning from examples as an inverse problem
- Learning Theory
- Nonparametric stochastic approximation with large step-sizes
- On Learning Vector-Valued Functions
- On the mathematical foundations of learning
- Online gradient descent learning algorithms
- Online Learning as Stochastic Approximation of Regularization Paths: Optimality and Almost-Sure Convergence
- Operator-valued kernels for learning from functional response data
- Optimal rates for the regularized least-squares algorithm
- Rates of convergence of randomized Kaczmarz algorithms in Hilbert spaces
- Sobolev norm learning rates for regularized least-squares algorithms
- Stochastic subspace correction in Hilbert space
- Support Vector Machines
- VECTOR VALUED REPRODUCING KERNEL HILBERT SPACES AND UNIVERSALITY
- VECTOR VALUED REPRODUCING KERNEL HILBERT SPACES OF INTEGRABLE FUNCTIONS AND MERCER THEOREM
This page was built for publication: Convergence analysis of online algorithms for vector-valued kernel regression
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7238004)
