Computational Limits of A Distributed Algorithm For Smoothing Spline
From MaRDI portal
Abstract: In this paper, we explore statistical versus computational trade-off to address a basic question in the application of a distributed algorithm: what is the minimal computational cost in obtaining statistical optimality? In smoothing spline setup, we observe a phase transition phenomenon for the number of deployed machines that ends up being a simple proxy for computing cost. Specifically, a sharp upper bound for the number of machines is established: when the number is below this bound, statistical optimality (in terms of nonparametric estimation or testing) is achievable; otherwise, statistical optimality becomes impossible. These sharp bounds partly capture intrinsic computational limits of the distributed algorithm considered in this paper, and turn out to be fully determined by the smoothness of the regression function. As a side remark, we argue that sample splitting may be viewed as an alternative form of regularization, playing a similar role as smoothing parameter.
Recommendations
- A distributed algorithm for H-infinity fixed-lag smoothing
- Fast computation of smoothing splines subject to equality constraints
- An algorithm for computing constrained smoothing spline functions
- A Distributed Algorithm for Least Squares Solutions
- Analysis of a class of parallel multigrid smoothers
- scientific article; zbMATH DE number 4049021
- A Parallel Algorithm for Mesh Smoothing
- On the use of smoothing to improve the performance of the splitting method
- Remarks on some parallel computation for spline recurrence formulas
Cites work
- A partially linear framework for massive heterogeneous data
- Asymptotically minimax hypothesis testing for nonparametric alternatives. III
- Divide and conquer kernel ridge regression: a distributed algorithm with minimax optimal rates
- scientific article; zbMATH DE number 45848 (Why is no real title available?)
- Local and global asymptotic inference in smoothing spline models
- Mathematical Statistics
- Maximum penalized likelihood estimation. Volume II: Regression
- Optimal estimation of the mean function based on discretely sampled functional data: phase transition
- Time series: theory and methods
Cited in
(36)- Distributed regression learning with coefficient regularization
- Circumventing superefficiency: an effective strategy for distributed computing in non-standard problems
- Surface temperature monitoring in liver procurement via functional variance change-point analysis
- Nonparametric distributed learning under general designs
- Computing confidence intervals from massive data via penalized quantile smoothing splines
- Divide-and-conquer information-based optimal subdata selection algorithm
- Divide and Recombine Approaches for Fitting Smoothing Spline Models with Large Datasets
- Distributed Generalized Cross-Validation for Divide-and-Conquer Kernel Ridge Regression and Its Asymptotic Optimality
- Harnessing Smoothness to Accelerate Distributed Optimization
- WONDER: weighted one-shot distributed ridge regression in high dimensions
- Distributed kernel ridge regression with communications
- scientific article; zbMATH DE number 7255125 (Why is no real title available?)
- Variance reduced median-of-means estimator for Byzantine-robust distributed inference
- Partitioned Approach for High-dimensional Confidence Intervals with Large Split Sizes
- Smoothing Splines Approximation Using Hilbert Curve Basis Selection
- Communication-Efficient Distributed Linear Discriminant Analysis for Binary Classification
- A review of distributed statistical inference
- Model aggregation for doubly divided data with large size and large dimension
- Distributed Bayesian inference in massive spatial data
- Distributed statistical inference for linear models with multi-source massive heterogeneous data
- A communication efficient distributed one-step estimation
- A distributed multiple sample testing for massive data
- Communication-Efficient Accurate Statistical Estimation
- Distributed adaptive nearest neighbor classifier: algorithm and theory
- Renewable composite quantile method and algorithm for nonparametric models with streaming data
- Adaptive distributed inference for multi-source massive heterogeneous data
- Distributed robust estimation and inference with contaminated data
- Grid Point Approximation for Distributed Nonparametric Smoothing and Prediction
- Divide and conquer for generalized approximately expectile regression
- Subsampling for big data linear models with measurement errors
- Efficient Nonparametric Estimation of 3D Point Cloud Signals through Distributed Learning
- Distributed Heterogeneity Learning for Generalized Partially Linear Models with Spatially Varying Coefficients
- Byzantine-Robust Distributed One-Step Estimation
- Efficient Quantization Mean Estimation for Distributed Learning
- Decentralized sparse linear regression via gradient-tracking
- A repeated block perturbation subsampling for large-scale longitudinal data
This page was built for publication: Computational Limits of A Distributed Algorithm For Smoothing Spline
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4637031)
