研究目的
To examine the issue of submodularity related to the optimization of sensor monitoring schemes using the value of information metric and to illustrate how greedy optimization approaches using value of information can lead to sub-optimal solutions for sensing in certain situations.
研究成果
The paper concludes that while greedy optimization approaches are efficient, they do not assess the value of multiple measurements taken together. The proposed heuristic approach combining the conditional entropy surrogate metric with VoI serves as an improvement over the standard forward greedy optimization using the VoI only.
研究不足
The lack of submodularity of the VoI metric can have a significant effect on the performance of a greedy optimization approach. The forward greedy approach can overspend on less informative single measurements while missing the potential for a greater return on investment from valuable pairs of measures.
1：Experimental Design and Method Selection:
The paper examines the use of an efficient, approximate 'greedy' optimization approach to solving the problem of sensor placement and scheduling. This approach iteratively adds or removes one measurement from a proposed scheme to maximize its net VoI.
2：Sample Selection and Data Sources:
The examples make use of Gaussian process spatio-temporal random field models to describe the variables which define system performance.
3：List of Experimental Equipment and Materials:
Not explicitly mentioned.
4：Experimental Procedures and Operational Workflow:
The paper presents algorithms for forward and reverse greedy optimization approaches and a heuristic approach combining the conditional entropy surrogate metric with VoI.
5：Data Analysis Methods:
The paper discusses the evaluation of the VoI and the use of conditional entropy as a submodular surrogate metric.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
