Skip to Main content Skip to Navigation
Journal articles

Sequential online subsampling for thinning experimental designs

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 µ, and only a given proportion α ∈ (0, 1) can be selected. The objective is to select good candidates X i on the y and maximize a concave function Φ of the corresponding information matrix. The optimal solution corresponds to the construction of an optimal bounded design measure ξ * α ≤ µ/α, with the diculty that µ is unknown and ξ * α must be constructed online. The construction proposed relies on the denition of a threshold τ on the directional derivative of Φ at the current information matrix, the value of τ being xed by a certain quantile of the distribution of this directional derivative. Combination with recursive quantile estimation yields a nonlinear two-timescale 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.
Document type :
Journal articles
Complete list of metadata

Cited literature [33 references]  Display  Hide  Download
Contributor : Luc Pronzato <>
Submitted on : Monday, August 3, 2020 - 9:39:31 AM
Last modification on : Friday, May 28, 2021 - 3:58:03 PM


Files produced by the author(s)




Luc Pronzato, Haiying Wang. Sequential online subsampling for thinning experimental designs. Journal of Statistical Planning and Inference, Elsevier, 2021, 212, pp.169-193. ⟨10.1016/j.jspi.2020.08.001⟩. ⟨hal-02527330v2⟩



Record views


Files downloads