Learning general sparse additive models from point queries in high dimensions
In learning methods, unknown functions \(f\) in multiple variables are to be recovered by a small number of samples taken from a compactum \(G\), say, where the functions are defined. In this paper, a probabilistic recovery of the unknown functions \(f\) is proposed for the compact set \(G\) being the multivariate unit-cube. The setting is so that the unknown function is additive (so-called SPAM, sparse additive models) with respect to its interactions. For this, the \(r\)-wise sets of (\(r\)-variate) interactions between the variables are used to form the sums that yield the unknown function \(f\). The goal is then to recover the sets of unknown interactions to form \(f\), where the recovery is to be understood in a probabilistic sense. The function \(f\) that underlies this approach is \textit{not} needed to be continuously differentiable (only Hölder continuous) as is required in other approaches. In the algorithms it is allowed to take samples within any points in \(G\) and the function values there at the discretion of the user. The \(r\)-wise interactions that are used are allowed for \(r\) \textit{arbitrary}, which is also new in this paper as compared to the known literature. The paper is organised such that the cases \(d=1\) and \(d=2\), uni- and bivariate, are always described first and then the new method is generalised to any dimension \(d\). An extensive discussion of the success of the new algorithms is provided at the end of the article.
- Algorithms for learning sparse additive models with interactions in high dimensions
- Learning sparse polynomial functions
- Capturing ridge functions in high dimensions from point queries
- Learning functions of few arbitrary linear parameters in high dimensions
- Learning non-parametric basis independent models from point queries via low-rank methods
- A Bound on Tail Probabilities for Quadratic Forms in Independent Random Variables
- A lasso for hierarchical interactions
- A mathematical introduction to compressive sensing
- A simple proof of the restricted isometry property for random matrices
- Algorithms for learning sparse additive models with interactions in high dimensions
- Approximation of functions of few variables in high dimensions
- Approximation of infinitely differentiable multivariate functions is intractable
- Component selection and smoothing in multivariate nonparametric regression
- Compressive sensing and structured random matrices
- Convergence rates of best \(N\)-term Galerkin approximations for a class of elliptic SPDEs
- Exact and Stable Covariance Estimation From Quadratic Sampling via Convex Programming
- Function spaces in Lipschitz domains and optimal rates of convergence for sampling
- High-dimensional additive modeling
- scientific article; zbMATH DE number 3898623 (Why is no real title available?)
- scientific article; zbMATH DE number 729680 (Why is no real title available?)
- Iterative hard thresholding for compressed sensing
- Learning juntas
- Minimax-optimal nonparametric regression in high dimensions
- Minimax-optimal rates for sparse additive models over kernel classes via convex programming
- Sampling numbers and function spaces
- Second-order cone programming
- Sparse additive models
- Sparse Volterra and Polynomial Regression Models: Recoverability and Estimation
- Sparsity in multiple kernel learning
- Statistical inference in compound functional models
- Surface estimation, variable selection, and the nonparametric oracle property
- Survey of modeling and optimization strategies to solve high-dimensional design problems with computationally-expensive black-box functions
- The restricted isometry property and its implications for compressed sensing
- Tight conditions for consistency of variable selection in the context of high dimensionality
- Variable selection in nonparametric additive models
- Variable selection using adaptive nonlinear interaction structures in high dimensions
- Variable selection with the strong heredity constraint and its oracle property
- Widths of embeddings in function spaces
- Learning a function from noisy samples at a finite sparse set of points
- Dimension-dependent error estimates for sampling recovery on Smolyak grids based on B-spline quasi-interpolation
- Sampling schemes and recovery algorithms for functions of few coordinate variables
- Algorithms for learning sparse additive models with interactions in high dimensions
- Towards optimal sampling for learning sparse approximation in high dimensions
- Learning non-parametric basis independent models from point queries via low-rank methods
This page was built for publication: Learning general sparse additive models from point queries in high dimensions
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2007617)
