Deep convolutional neural networks (DCNNs) have shown outstanding performance in the fields of computer vision, natural language processing, and complex system analysis. With the improvement of performance with deeper layers, DCNNs incur higher computational complexity and larger storage requirement, making it extremely difficult to deploy DCNNs on resource-limited embedded systems (such as mobile devices or Internet of Things devices). Network quantization efficiently reduces storage space required by DCNNs. However, the performance of DCNNs often drops rapidly as the quantization bit reduces. In this article, we propose a space efficient quantization scheme which uses eight or less bits to represent the original 32-bit weights. We adopt singular value decomposition (SVD) method to decrease the parameter size of fully-connected layers for further compression. Additionally, we propose a weight clipping method based on dynamic boundary to improve the performance when using lower precision. Experimental results demonstrate that our approach can achieve up to approximately 14x compression while preserving almost the same accuracy compared with the full-precision models. The proposed weight clipping method can also significantly improve the performance of DCNNs when lower precision is required.

Mobile social sensing network is one kind of emerging networks in which sensing tasks are performed by mobile users and sensing data are shared and collected by leveraging the intermittent inter-contacts among mobile users. Traditional ad hoc routing protocols are inapplicable or perform poorly for data collection or data sharing in such mobile social networks because nodes are seldom fully connected. In recent years, many routing protocols (especially social-based routing) are proposed to improve the delivery ratio in mobile social networks, but most of them do not consider the load of nodes thus may lead to unbalanced energy consumption among nodes. In this paper, we propose a simple Energy Efficient framework for Social-based Routing (EE-SR) in mobile social sensing networks to balance the load of nodes while maintaining the delivery ratio within an acceptable range by limiting the chances of forwarding in traditional social-based routing. Furthermore, we also propose an improved version of EE-SR to dynamically adjust the controlling parameter. Simulation results on real-life mobile traces demonstrate the efficiency of our proposed framework.

Location privacy has been a serious concern for mobile users who use location-based services provided by third-party providers via mobile networks. Recently, there have been tremendous efforts on developing new anonymity or obfuscation techniques to protect location privacy of mobile users. Though effective in certain scenarios, these existing techniques usually assume that a user has a constant privacy requirement along spatial and/or temporal dimensions, which may be not true in real-life scenarios. In this paper, we introduce a new location privacy problem: Location-aware Location Privacy Protection (L2P2) problem, where users can define dynamic and diverse privacy requirements for different locations. The goal of the L2P2 problem is to find the smallest cloaking area for each location request so that diverse privacy requirements over spatial and/or temporal dimensions are satisfied for each user. In this paper, we formalize two versions of the L2P2 problem, and propose several efficient heuristics to provide such location-aware location privacy protection for mobile users. Through extensive simulations over large synthetic and real-life datasets, we confirm the effectiveness and efficiency of the proposed L2P2 algorithms.

Shortest path routing protocol intends to minimize the total delay between every pair of destination node and source node. However, it is also well-known that shortest path routing suffers from uneven distribution of traffic load, especially in dense wireless networks. Recently, several new routing protocols are proposed in order to balance traffic load among nodes in a network. One of them is Circular Sailing Routing (CSR) which maps nodes on the surface of a sphere and select routes based on surface distances. CSR has been demonstrated with better load balance than shortest path routing via simulations. However, it is still open that what load distribution CSR can achieve. Therefore, in this paper, we theoretically analyze the traffic load distribution of CSR in a dense circular wireless network. Using the techniques developed by Hyttiä and Virtamo, we are able to derive the traffic load of any point inside the network. We then conduct extensive simulations to verify our theoretical results with grid and random networks.