Sequential online subsampling for thinning experimental designs

DOI10.1016/J.JSPI.2020.08.001zbMATH Open1460.62128arXiv2004.00792OpenAlexW3047634786MaRDI QIDQ830692FDOQ830692

Luc Pronzato, Haiying Wang

Publication date: 7 May 2021

Published in: Journal of Statistical Planning and Inference (Search for Journal in Brave)

Abstract: We consider a design problem where experimental conditions (design points $X_i$) are presented in the form of a sequence of i.i.d. random variables, generated with an unknown probability measure $mu$, and only a given proportion $alphain(0,1)$ can be selected. The objective is to select good candidates $X_i$ on the fly and maximize a concave function $Phi$ of the corresponding information matrix. The optimal solution corresponds to the construction of an optimal bounded design measure $x{i}_{a}lph{a}^{*}leqmu/alpha$, with the difficulty that $mu$ is unknown and $x{i}_{a}lph{a}^{*}$ must be constructed online. The construction proposed relies on the definition of a threshold $au$ on the directional derivative of $Phi$ at the current information matrix, the value of $au$ being fixed by a certain quantile of the distribution of this directional derivative. Combination with recursive quantile estimation yields a nonlinear two-time-scale stochastic approximation method. It can be applied to very long design sequences since only the current information matrix and estimated quantile need to be stored. Convergence to an optimum design is proved. Various illustrative examples are presented.

Full work available at URL: https://arxiv.org/abs/2004.00792

Recommendations

• Sequential construction of an experimental design from an i.i.d. sequence of experiments without replacement
• On the sequential construction of optimum bounded designs
• Sequential experimental design and response optimisation
• scientific article; zbMATH DE number 2172867
• scientific article; zbMATH DE number 1747152

zbMATH Keywords

sequential designdesign of experimentssubsamplingactive learningdata thinning

Mathematics Subject Classification ID

Optimal statistical designs (62K05) Sequential statistical design (62L05) Stochastic approximation (62L20) Computational learning theory (68Q32)

Cites Work

• Asymptotic Statistics
• Title not available (Why is that?)
• A Space-Efficient Recursive Procedure for Estimating a Quantile of an Unknown Distribution
• Optimal design with bounded density: Optimization algorithms of the exchange type
• Title not available (Why is that?)
• An Algorithm for the Construction of "D-Optimal" Experimental Designs
• Analysis of recursive stochastic algorithms
• Title not available (Why is that?)
• Some remarks on the value-at-risk and the conditional value-at-risk
• Title not available (Why is that?)
• Model-oriented design of experiments
• Stochastic approximation with two time scales
• A Comparison of Algorithms for Constructing Exact D-Optimal Designs
• Asymptotic Optimal Policies for the Stochastic Sequential Assignment Problem
• Title not available (Why is that?)
• Design of experiments in nonlinear models. Asymptotic normality, optimality criteria and small-sample properties
• On the sequential construction of optimum bounded designs
• Convergence rate of linear two-time-scale stochastic approximation.
• Accelerated Stochastic Approximation
• A Note on Optimal Bounded Designs
• Information-regret compromise in covariate-adaptive treatment allocation
• Title not available (Why is that?)
• Information-Based Optimal Subdata Selection for Big Data Linear Regression
• A minimax equivalence theorem for optimum bounded design measures
• A stability criterion for two timescale stochastic approximation schemes
• Optimality in identification of linear plants
• Optimal and asymptotically optimal decision rules for sequential screening and resource allocation
• Minimax robust active learning for approximately specified regression models

Cited In (6)

• Online Selection of Alternating Subsequences from a Random Sample
• Optimal sampling designs for multidimensional streaming time series with application to power grid sensor data
• On the sequential construction of optimum bounded designs
• Optimal subsampling design for polynomial regression in one covariate
• Unweighted estimation based on optimal sample under measurement constraints
• A note on centering in subsample selection for linear regression

This page was built for publication: Sequential online subsampling for thinning experimental designs

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q830692)