Search

Abstract: In wireless sensor networks, individual sensing values are not reliable due to node failures. The effect of these failures can be reduced by using aggregated values for groups of sensor nodes instead of the individual sensing values. However, most existing works have focused on computing the aggregation of all the nodes without grouping. Only a few approaches dealt with the processing of grouped aggregate queries. However, since groups in their approaches are disjoint, some areas which are not covered by groups cannot be considered, even if the areas are relevant to the user’s interest. In this paper, we propose a new type of queries, region-based queries, and a framework to process region-based queries, called REQUEST +. A region in REQUEST+ is defined as a maximal set of nodes located within a circle having a diameter specified in the query. To efficiently construct a large number of regions covering the entire monitoring area, we build the SEC (Smallest Enclosing Circle) index. Moreover, in order to process a region-based query, we adapt a clustering-based aggregation method, in which there is a leader node for each region. To minimize the communication cost, we formulate an optimal leader selection problem and prove that it is NP-hard. In addition, we transform the problem into the weighted set-cover problem to utilize the algorithm devised for the problem. Finally, we construct a query-initiated routing tree for the communication between the leader and non-leader nodes. In the experimental results, we show that the result of our region-based query is more reliable than that of the query which is based on individual nodes, and our processing method is more energy-efficient than existing methods for processing grouped aggregate queries. [Copyright &y& Elsevier]

Despite the active research on deep learning these days, no existing works attempt to query neural networks for a specific task so that we can quickly obtain a model specialized for the queried task. This paper presents the first study on the problem of querying neural networks, aiming to efficiently find a lightweight model for any on-demand sub-task supported by a large and generic neural network. This problem is well motivated by the fact that such a lightweight model is particularly suitable for deployment in commodity mobile devices with limited computing resources. In this paper, we propose QueryNet , a framework of queriable neural networks, which is based on our class-aware channel pruning technique and the proposed method of merging multiple tiny networks for producing a specialized model for the task. Through the extensive experiments, we show that QueryNet can generate a lightweight neural network for a given task up to 3.25 times faster than learning a specialized model from scratch, and the resulting specialized neural network carries up to 37 times less parameters than a pretrained network, and even shows a slightly better accuracy than that of the pretrained model. [ABSTRACT FROM AUTHOR]