Recommendation systems suggest items to users by predicting their preferences based on historical data. The industry traditionally handles large-scale recommendation requests by scaling the number of devices without much concern for a single device’s performance. However, there is a trend for recommendation systems to gradually move from a centralized service to an edge device. The edge-scale recommendation systems have distinct features that are different from traditional large-scale deployments, which poses different challenges to the acceleration of the recommendation system. In this paper, we focus on the edge-scale recommendation system and propose an inter-stage parallelism optimization method deployed on a single GPU. Experiments show that our framework could improve recommendation system throughput by 1.89×~2.2× for different datasets on the GPU.

Spectrum sensing is one of the main components of the cognitive radio (CR) system, based on which secondary users (SUs) can get access to spectrum holes available. On the other hand, a malicious adversary can also attack the primary user (PU) system and the legitimate CR system via spectrum sensing, which can lead to serious security issue for both systems. In this article, we study an online attacking strategy, referred to as PU emulation attacker (PEA), which transmits forged PU signals over available channels to deteriorate the spectrum sensing performance of the SUs. We propose an online learning based attacking scheme for both the single attacker and the multiple-attacker cases, and analyze the regret upper bound of the proposed algorithm. The proposed PEA strategy can work in both stationary and non-stationary CR networks where the statistical characteristics of channels and the access strategy of SUs change over time. Numerical results show that it is more efficient than others in two different performance metrics: successful accesses of SUs and effective attacks of PEAs. Our proposal raises an interesting open question on how to develop CR networks with security guarantee.

Efficient probing spectrum holes is one of the most challenging tasks for the secondary user (SU) in a cognitive radio (CR) network. In this paper, we introduce a novel spectrum sensing framework where the duration for sensing at each time slot is variable. Sensing more channels increases the probability of finding a spectral hole, however, it would spend more time for sensing inevitably, which reduces the time for data transmission at a given time slot. Considering the sensing-transmission tradeoff, the optimization goal of spectrum sensing strategy is set to maximize the expected achievable throughput of the SU, which is formulated as a partially observable Markov decision process (POMDP). Finding an optimal solution to this optimization problem is computationally expensive due to its large state space, as well as large action space. We develop a novel spectrum sensing strategy based on deep reinforcement learning, which converges fast and can deal with complex scenario. Numerical results show that our proposed strategy can improve system throughput significantly.

In this paper, we propose to learn Lifestyles of mobile users via mobile Phone Sensing (LIPS), and we develop a system and algorithms to realize this idea. First, we present the workflow and architecture of our system, LIPS. Combining both unsupervised and supervised learning, we propose a hybrid scheme for lifestyle learning, which consists of two parts: characterization and prediction. Specifically, we present a two-stage algorithm to characterize the lifestyle of a mobile user using Places of Interest (PoIs), which leverages two different algorithms for coarse-grained and fine-grained clustering in two stages respectively. Based on discovered PoIs, we present a method to build a model to predict his/her future activities using a supervised classification algorithm. In addition, we present an adaptive sampling algorithm for improving energy efficiency, which leverages both the discovered PoIs and the lifestyle model for adaptively controlling the sampling rate. We implemented the proposed system and algorithms based on the Android platform. We have validated and evaluated LIPS via extensive field tests carried out for over 1.5 months in 6 cities of USA. The experimental results show that LIPS can 1) well discover PoIs of mobile users, 2) precisely predict their future activities, and 3) achieve significant energy savings (compared to periodic sampling).

Device-to-device (D2D) communication has emerged as a promising technique for improving capacity and reducing power consumption in wireless networks. Most existing works on D2D communications either targeted CDMA-based single-channel networks or aimed at maximizing network throughput. In this paper, we, however, aim to enable green D2D communications in OFDMA-based wireless networks. We formally define an optimization problem based on a practical link data rate model, whose objective is to minimize total power consumption while meeting user data rate requirements. We propose solving it using a joint optimization approach by presenting two effective and efficient algorithms, which both jointly determines mode selection, channel allocation and power assignment. It has been shown by extensive simulation results that the proposed algorithms can achieve over 68% power savings, compared to several baseline methods.

In a mobile crowd sensing system, a smartphone undertakes many different sensing tasks that demand data from various sensors. In this paper, we consider the problem of scheduling different sensing tasks assigned to a smartphone with the objective of minimizing sensing energy consumption while ensuring Quality of SenSing (QoSS). First, we consider a simple case in which each sensing task only requests data from a single sensor. We formally define the corresponding problem as the Minimum Energy Single-sensor task Scheduling (MESS) problem and present a polynomial-time optimal algorithm to solve it. Furthermore, we address a more general case in which some sensing tasks request multiple sensors to report their measurements simultaneously. We present an Integer Linear Programming (ILP) formulation as well as an effective polynomial-time heuristic algorithm, for the corresponding Minimum Energy Multi-sensor task Scheduling (MEMS) problem. Extensive simulation results show that the proposed algorithms achieve over 79% energy savings on average compared to a widely-used baseline approach, and moreover, the proposed heuristic algorithm produces close-to-optimal solutions.

With the exploding popularity of wireless communication, the radio spectrum has become a scarce commodity. To further improve the network capacity, various solutions have been proposed to increase spectrum efficiency and network throughput. Small cell network is one of these new trends for next generation mobile network design. One model is using relay stations (RS) as small cell providers to achieve extended coverage, lower cost, and higher network capacity. Considering multiple related physical constraints such as channel capacity, signal to noise ratio (SNR) requirement of subscribers, relay power and network topology, this paper studies a joint signal-aware RS placement and power allocation problem with multiple base stations in wireless relay networks. We presented approximation schemes which first find a minimum number of RS, using maximum transmission power, to cover all the subscribers meeting each SNR requirement, and then ensure communications between any subscriber to a base station by adjusting the transmission power of each RS. Numerical results are presented to confirm the theoretical analysis of our schemes, and to show strong performances of our solutions.

Device-to-Device (D2D) communication has emerged as a promising technique for improving capacity and reducing power consumption in wireless networks. Most existing works on D2D communications either targeted CDMA-based single-channel networks or aimed to maximize network throughput. In this paper, we, however, aim at enabling green D2D communications in OFDMA-based wireless networks. We formally define an optimization problem based on a practical link data rate model, whose objective is to minimize power consumption while meeting user data rate requirements. We then present an effective algorithm to solve it in polynomial time, which jointly determines mode selection, channel allocation and power assignment. It has been shown by extensive simulation results that the proposed algorithm can achieve over 57% power savings, compared to several baseline methods.

Currently, a few online review and recommendation systems (such as Yelp and Trip Advisor) have attracted millions of users and are gaining increasing popularity. They usually rate and rank places and attractions based on subjective ratings provided by users. In this paper, we present design, implementation and evaluation of a mobile phone Sensing based Objective Ranking (SOR) system, which ranks a target place based on data collected via mobile phone sensing. Our system has the following desirable features: 1) it is easy to use, 2) its architecture is so scalable that various embedded and external sensors can be easily integrated into it, 3) an online scheduling algorithm is proposed and used to schedule sensing activities for coverage maximization, which has a constant approximation ratio of 1/2, 4) a personalizable ranking algorithm is developed and used to rank target places based on various sensor readings and user preferences. We validate and evaluate SOR via both field tests (using real hiking trails and coffee shops in Syracuse, NY as target places) and simulation. The field-testing results show that data collected and processed by SOR can well capture characteristics of target places, and personalizable rankings produced by SOR can well match user preferences. In addition, simulation results well justify effectiveness of the proposed scheduling algorithm.

In this paper, we consider leveraging GPS-less energy-efficient sensing scheduling for mobile crowd sensing. We present a probabilistic model for sensing coverage without accurate location information (provided by GPS), based on which we formally define the Energy-constrained Maximum Coverage Sensing Scheduling (E-MCSS) problem for maximum coverage and the Fair Maximum Coverage Sensing Scheduling (F-MCSS) problem for fairness. Assuming that moving trajectories of mobile users are known beforehand, we present a (1 - 1/e)-approximation algorithm and a 1/2-approximation algorithm to solve the E-MCSS and F-MCSS problems in polynomial time, respectively, which can serve as benchmarks for performance evaluation. Under realistic assumptions, we present a GPS-less energy-efficient protocol for sensing scheduling based on the proposed algorithms. We developed an Android-based mobile crowd sensing system, on which we implemented the proposed protocol. Simulation results and experimental results (from a field test) are presented to validate and justify effectiveness of the proposed algorithms and protocol.

Small cell network is the new trend for next generation mobile network design. One feasible model is using Relay stations (RS) as small cell providers to achieve extended coverage, lower cost, and higher network capacity. This paper studies Signal-aware relay station placement and power allocation problem in wireless relay networks with multiple base stations in the field. This problem consists of both subscriber coverage problem and relay power optimization problem, which have not been extensively studied together in previous works. This work takes into account physical constraints such as channel capacity, signal to noise ratio (SNR) requirement of subscribers, relay power cost and network topology. We set up a two-step goal that is firstly to find minimum number of RS in order to cover all the subscribers meeting each SNR requirement, and then to ensure communications built between any subscriber to a base station. In order to ensure each subscriber's SNR, transmission power of each RS should be adjustable. Thus, minimizing power cost of RSs is our goal in the second step. We divide the problem into two sub-problems, Lower-tier Coverage Relay Allocation (LCRA) problem and Upper-tier Connectivity Relay Allocation (UCRA) problem. For the LCRA problem, we present two approximation solutions based on minimum hitting set and maximum independent set. For the UCRA problem, an approximation algorithm and an optimal algorithm are proposed. At the end, an approximation solution for our original problem, which combines the approaches of the two sub-problems, is provided. Numerical results are presented to confirm the theoretical analysis of our schemes, and to show strong performances of our solutions.

With wireless resource virtualization, multiple Mobile Virtual Network Operators (MVNOs) can be supported over a shared physical wireless network and traffic loads in a Base Station (BS) can be easily migrated to more power-efficient BSs in its neighborhood such that idle BSs can be turned off or put into sleep to save power. In this paper, we propose to leverage load migration and BS consolidation for green communications and consider a power-efficient network planning problem in virtualized Cognitive Radio Networks (CRNs) with the objective of minimizing total power consumption while meeting traffic load demand of each MVNO. First, we present a Mixed Integer Linear Programming (MILP) to provide optimal solutions. Then we present a general optimization framework to guide algorithm design, which solves two subproblems, channel assignment and load allocation, in sequence. For channel assignment, we present a (Δ1)-approximation algorithm (where Δ is the maximum number of BSs a BS can potentially interfere with). For load allocation, we present a polynomial-time optimal algorithm for a special case where BSs are power-proportional as well as two effective heuristic algorithms for the general case. In addition, we present an effective heuristic algorithm that jointly solves the two subproblems. It has been shown by extensive simulation results that the proposed algorithms produce close-to-optimal solutions, and moreover, achieve over 45% power savings compared to a baseline algorithm that does not migrate loads or consolidate BSs.

Mobile phones with a rich set of embedded sensors enable sensing applications in various domains. In this paper, we propose to leverage cloud-assisted collaborative sensing to reduce sensing energy consumption for mobile phone sensing applications. We formally define a minimum energy sensing scheduling problem and present a polynomial-time algorithm to obtain optimal solutions, which can be used to show energy savings that can potentially be achieved by using collaborative sensing in mobile phone sensing applications, and can also serve as a benchmark for performance evaluation. We also address individual energy consumption and fairness by presenting an algorithm to find fair energy-efficient sensing schedules. Under realistic assumptions, we present two practical and effective heuristic algorithms to find energy-efficient sensing schedules. It has been shown by simulation results based on real energy consumption (measured by the Monsoon power monitor) and location (collected from the Google Map) data that collaborative sensing significantly reduces energy consumption compared to a traditional approach without collaborations, and the proposed heuristic algorithm performs well in terms of both total energy consumption and fairness.