Quantization for Maximin ARE in Distributed Estimation

We consider the design of optimal quantizers for the distributed estimation of a deterministic parameter. In particular, we design deterministic scalar quantizers to maximize the minimum asymptotic relative efficiency (ARE) between quantized and unquantized ML estimators. We first design identical quantizers using the class of score-function quantizers (SFQ). We show that the structure of SFQs generally depend on the parameter value, but can be expressed as thresholds on the sufficient statistic for a large class of distributions. We provide a convergent iterative algorithm to obtain the best SFQ that maximizes the minimum ARE for distributions of that class. We compare the performance of the optimal SFQ with a general quantizer designed without making any restrictions on the structure. This general quantizer is hard to implement due to lack of structure, but is optimal if the iterative design algorithm does not encounter local minima. Through numerical simulations, we illustrate that the two quantizers designed are identical. In other words, the optimal quantizer structure is that of an SFQ. For a distributed estimation setup, designing identical quantizers is shown to be suboptimal. We, therefore, propose a joint multiple quantizer design algorithm based on a person-by-person optimization technique employing the SFQ structure. Using numerical examples, we illustrate the gain in performance due to designing nonidentical quantizers.

[1]  Amy R. Reibman,et al.  Design of quantizers for decentralized estimation systems , 1993, IEEE Trans. Commun..

[2]  Toby Berger,et al.  Estimation via compressed information , 1988, IEEE Trans. Inf. Theory.

[3]  A. Swami,et al.  Quantization For Distributed Estimation in Large Scale Sensor Networks , 2005, 2005 3rd International Conference on Intelligent Sensing and Information Processing.

[4]  R. Radner,et al.  Team Decision Problems , 1962 .

[5]  A. Swami,et al.  Score-Function Quantization for Distributed Estimation , 2006, 2006 40th Annual Conference on Information Sciences and Systems.

[6]  John N. Tsitsiklis,et al.  Extremal properties of likelihood-ratio quantizers , 1993, IEEE Trans. Commun..

[7]  John N. Tsitsiklis,et al.  Decentralized detection by a large number of sensors , 1988, Math. Control. Signals Syst..

[8]  H. Vincent Poor,et al.  An Introduction to Signal Detection and Estimation , 1994, Springer Texts in Electrical Engineering.

[9]  Alejandro Ribeiro,et al.  Non-parametric distributed quantization-estimation using wireless sensor networks , 2005, Proceedings. (ICASSP '05). IEEE International Conference on Acoustics, Speech, and Signal Processing, 2005..

[10]  G. Casella,et al.  Statistical Inference , 2003, Encyclopedia of Social Network Analysis and Mining.

[11]  X.R. Li,et al.  Optimal sensor data quantization for best linear unbiased estimation fusion , 2004, 2004 43rd IEEE Conference on Decision and Control (CDC) (IEEE Cat. No.04CH37601).

[12]  M.L. Fowler,et al.  Fisher-information-based data compression for estimation using two sensors , 2005, IEEE Transactions on Aerospace and Electronic Systems.

[13]  Shun-ichi Amari,et al.  Parameter estimation with multiterminal data compression , 1995, IEEE Trans. Inf. Theory.

[14]  John A. Gubner,et al.  Distributed estimation and quantization , 1993, IEEE Trans. Inf. Theory.

[15]  Zhi-Quan Luo,et al.  Universal decentralized estimation in a bandwidth constrained sensor network , 2005, IEEE Transactions on Information Theory.