We study randomized experiments in a service system when stochastic congestion can arise from temporarily limited supply or excess demand. Such congestion gives rise to cross-unit interference between the waiting customers, and analytic strategies that do not account for this interference may be biased. In current practice, one of the most widely used ways to address stochastic congestion is to use switchback experiments that alternatively turn a target intervention on and off for the whole system. We find, however, that under a queueing model for stochastic congestion, the standard way of analyzing switchbacks is inefficient, and that estimators that leverage the queueing model can be materially more accurate. We also consider a new experimental design, which we refer to as the length-0 switchback, that can be used to estimate a policy gradient of the dynamic system using only unit-level randomization. This design avoids needing to pre-commit to a switchback length before data collection, and can thus be easier to deploy in settings with nonstationarity.

Thompson sampling has proven effective across a wide range of stationary bandit environments. However, as we demonstrate in this paper, it can perform poorly when applied to nonstationary environments. We show that such failures are attributed to the fact that, when exploring, the algorithm does not differentiate actions based on how quickly the information acquired loses its usefulness due to nonstationarity. Building upon this insight, we propose predictive sampling, an algorithm that deprioritizes acquiring information that quickly loses usefulness. Theoretical guarantee on the performance of predictive sampling is established through a Bayesian regret bound. We provide versions of predictive sampling for which computations tractably scale to complex bandit environments of practical interest. Through numerical simulation, we demonstrate that predictive sampling outperforms Thompson sampling in all nonstationary environments examined.

Convex optimization with feedback is a framework where a learner relies on iterative queries and feedback to arrive at the minimizer of a convex function. It has gained considerable popularity thanks to its scalability in large-scale optimization and machine learning. The repeated interactions, however, expose the learner to privacy risks from eavesdropping adversaries that observe the submitted queries. In this paper, we study how to optimally obfuscate the learner’s queries in convex optimization with first-order feedback, so that their learned optimal value is provably difficult to estimate for an eavesdropping adversary. We consider two formulations of learner privacy: a Bayesian formulation in which the convex function is drawn randomly, and a maximin formulation in which the function is fixed and the adversary’s probability of error is measured with respect to a minimax criterion. Suppose that the learner wishes to ensure the adversary cannot estimate accurately with probability greater than$1/L$for some$L > 0$. Our main results show that the query complexity overhead is additive in$L$in the maximin formulation, but multiplicative in$L$in the Bayesian formulation. Compared to existing learner-private sequential learning models with binary feedback, our results apply to the significantly richer family of general convex functions with full-gradient feedback. Our proofs rely on tools from the theory of Dirichlet processes, as well as a novel strategy designed for measuring information leakage under a full-gradient oracle.

We study the query complexity of a learner-private sequential learning problem, motivated by the privacy and security concerns due to eavesdropping that arise in practical applications such as pricing and Federated Learning. A learner tries to estimate an unknown scalar value, by sequentially querying an external database and receiving binary responses; meanwhile, a third-party adversary observes the learner’s queries but not the responses. The learner’s goal is to design a querying strategy with the minimum number of queries (optimal query complexity) so that she can accurately estimate the true value, while the eavesdropping adversary even with the complete knowledge of her querying strategy cannot.

Convex optimization with feedback is a framework where a learner relies on iterative queries and feedback to arrive at the minimizer of a convex function. The paradigm has gained significant popularity recently thanks to its scalability in large-scale optimization and machine learning. The repeated interactions, however, expose the learner to privacy risks from eavesdropping adversaries that observe the submitted queries. In this paper, we study how to optimally obfuscate the learner’s queries in convex optimization with first-order feedback, so that their learned optimal value is provably difficult to estimate for the eavesdropping adversary. We consider two formulations of learner privacy: a Bayesian formulation in which the convex function is drawn randomly, and a minimax formulation in which the function is fixed and the adversary’s probability of error is measured with respect to a minimax criterion. We show that, if the learner wants to ensure the probability of the adversary estimating accurately be kept below 1/L, then the overhead in query complexity is additive in L in the minimax formulation, but multiplicative in L in the Bayesian formulation. Compared to existing learner-private sequential learning models with binary feedback, our results apply to the significantly richer family of general convex functions with full-gradient feedback. Our proofs are largely enabled by tools from the theory of Dirichlet processes, as well as more sophisticated lines of analysis aimed at measuring the amount of information leakage under a full-gradient oracle.

We formulate a private learning model to study an intrinsic tradeoff between privacy and query complexity in sequential learning. Our model involves a learner who aims to determine a scalar value, $v^*$, by sequentially querying an external database and receiving binary responses. In the meantime, an adversary observes the learner’s queries, though not the responses, and tries to infer from them the value of $v^*$. The objective of the learner is to obtain an accurate estimate of $v^*$ using only a small number of queries, while simultaneously protecting her privacy by making $v^*$ provably difficult to learn for the adversary. Our main results provide tight upper and lower bounds on the learner’s query complexity as a function of desired levels of privacy and estimation accuracy. We also construct explicit query strategies whose complexity is optimal up to an additive constant.

We study the query complexity of Bayesian Private Learning: a learner wishes to locate a random target within an interval by submitting queries, in the presence of an adversary who observes all of her queries but not the responses. How many queries are necessary and sufficient in order for the learner to accurately estimate the target, while simultaneously concealing the target from the adversary? Our main result is a query complexity lower bound that is tight up to the first order. We show that if the learner wants to estimate the target within an error of $\epsilon$, while ensuring that no adversary estimator can achieve a constant additive error with probability greater than $1/L$, then the query complexity is on the order of $L\log(1/\epsilon)$ as $\epsilon \to 0$. Our result demonstrates that increased privacy, as captured by $L$, comes at the expense of a \emph{multiplicative} increase in query complexity. The proof builds on Fano's inequality and properties of certain proportional-sampling estimators.

In this paper, we present the first detailed analysis of influence maximization in noncooperative social networks under the Independent Cascade Model (ICM). We propose a new influence model based on the ICM and prove the approximation guarantees for influence maximization in noncooperative settings. We structure the influence diffusion into two stages, namely, seed node selection and influence diffusion. In the former, we introduce a modified hierarchy-based seed node selection strategy, which can take node noncooperation into consideration. In the latter, we propose a game-theoretic model to characterize the behavior of noncooperative nodes and design a Vickrey- Clarke-Groves (VCG)-like scheme to incentivise cooperation. Then, we study the budget allocation problem between the two stages, and show that a marketer can utilize the two proposed strategies to tackle noncooperation intelligently. We evaluate our proposed schemes on large coauthorship networks, and the results show that our seed node selection scheme is very robust to noncooperation and the VCG-like scheme can effectively stimulate a node to become cooperative.

We propose and analyze a family of \emphinformation processing systems, where a finite set of experts or servers are employed to extract information about a stream of incoming jobs. Each job is associated with a hidden label drawn from some prior distribution. An inspection by an expert produces a noisy outcome that depends both on the job’s hidden label and the type of the expert, and occupies the expert for a finite time duration. A decision maker’s task is to dynamically assign inspections so that the resulting outcomes can be used to accurately recover the labels of all jobs, while keeping the system stable. Among our chief motivations are applications in crowd-sourcing, diagnostics, and experiment designs, where one wishes to efficiently discover the nature of a large number of items, using a finite pool of computational resources or human agents. We focus on the \emphcapacity of such an information processing system. Given a level of accuracy guarantee, we ask how many experts are needed in order to stabilize the system, and through what inspection architecture. Our main result provides an adaptive inspection policy that is asymptotically optimal in the following sense: the ratio between the required number of experts under our policy and the theoretical optimal converges to one, as the probability of error in label recovery tends to zero.

Characterizing the video diffusion in online social networks (OSNs) is not only instructive for network traffic engineering, but also provides insights into the information diffusion process. A number of continuous-time diffusion models have been proposed to describe video diffusion under the assumption that the activation latency along social links follows a single parametric distribution. However, such assumption has not been empirically verified. Moreover, a user usually has multiple activated neighbors with different activation times, and it is hard to distinguish the different contributions of these multiple potential sources. To fill this gap, we study the multiple-source-driven asynchronous information diffusion problem based on substantial video diffusion traces. Specifically, we first investigate the latency of information propagation along social links and define the single-source (SS) activation latency for an OSN user. We find that the SS activation latency follows the exponential mixture model. Then we develop an analytical framework which incorporates the temporal factor and the influence of multiple sources to describe the influence propagation process. We show that one's activation probability decreases exponentially with time. We also show that the time shift of the exponential function is only determined by the most recent source (MRS) active user, but the total activation probability is the combination of influence exerted by all active neighbors. Based on these discoveries, we develop a multi-source-driven asynchronous diffusion model (MADM). Using maximum likelihood techniques, we develop an algorithm based on expectation maximization (EM) to learn model parameters, and validate our proposed model with real data. The experimental results show that the MADM obtains better prediction accuracy under various evaluation metrics.

We study a multi-server model with n flexible servers and rn queues, connected through a fixed bipartite graph, where the level of flexibility is captured by the average degree, d(n), of the queues. Applications in content replication in data centers, skill-based routing in call centers, and flexible supply chains are among our main motivations. We focus on the scaling regime where the system size n tends to infinity, while the overall traffic intensity stays fixed. We show that a large capacity region (robustness) and diminishing queueing delay (performance) are jointly achievable even under very limited flexibility (d(n) l n). In particular, when d(n) gg ln n , a family of random-graph-based interconnection topologies is (with high probability) capable of stabilizing all admissible arrival rate vectors (under a bounded support assumption), while simultaneously ensuring a diminishing queueing delay, of order ln n/ d(n), as n-> ∞. Our analysis is centered around a new class of virtual-queue-based scheduling policies that rely on dynamically constructed partial matchings on the connectivity graph.

In this paper we analyze influence maximization for noncooperative social networks under the Independent Cascade Model. We propose a model of noncooperative nodes and prove some interesting properties of this model. Based on this, we further develop a game-theoretic model to characterize the behavior of noncooperative nodes, and design a Vickrey-Clarke-Groves-like scheme to incentivise cooperation. An advertiser can resolve the negative effect of noncooperation with our proposed solution. Evaluation on large social networks demonstrates the importance of cooperation and the effectiveness of our proposed incentive scheme in maximizing influence. We also discuss the budget allocation between seed nodes activation and incentives to non-seed nodes.

The rapid development of online social networks (OSN) renders them a popular mechanism for information diffusion. Studying the temporal characteristics is critical in understanding the diffusion process. However, due to the lack of well-defined propagation data, hardly any study addresses the temporal feature of information diffusion in OSN. In this paper, we present a measurement study on information diffusion in the Renren social network. We investigate the latency of information propagation along social links and define the “activation time” for an OSN user, and find that the activation time follows the lognormal distribution. Based on this, we develop two new information diffusion models incorporating asynchronous activation times. Application of the models in the influence maximization problem shows that they capture the temporal diffusion behavior very well. This leads to fundamental ramifications to many related OSN applications.

In this paper, we consider the problem of maximizing information propagation with noncooperative nodes in social networks. We generalize the linear threshold model to take node noncooperation into consideration and provide a provable approximation guarantees for the noncooperative influence maximization problem. We propose an analytical model based on the generalized maximum flow problem to characterize the noncooperative behavior of an individual node in maximizing influence. Based on this, we develop a new seed node selection strategy, under the linear threshold model, to account for user noncooperativeness. Extensive simulations on large collaboration networks show that our proposed flow-based strategy outperforms the weighted degree scheme under various noncooperative scenarios. The evaluation also validates the importance of cooperation and incentives in maximizing influence.

Social-based forwarding algorithms provide a new perspective on the study of routing in opportunistic mobile networks, and all of these schemes assume a uniform pattern for message generating rule. However, this is unconvincing due to the heterogeneity of contact rates in human communication patterns. In this paper we propose three social-based communication pattern models and utilize them to evaluate the network performance of different social-based routing protocols based on several human mobility traces. We find that communication patterns could significantly affect the network performance and the influence degree largely depends on the social metrics which these communication patterns are based on. We contend that considering communication pattern is quite important for designing a practical routing algorithm in opportunistic mobile networks.

In this paper, we study the problem of discovering multiple resource holders and how to evaluate a node's satisfaction in query incentive networks. Utilizing an acyclic tree, we show that query propagation has a nature of exponential start, polynomial growth, and eventually becoming a constant. We model the query propagation as an extensive game, obtain nodes' greedy behaviors from Nash equilibrium analysis, and show the impairment of greedy behaviors via a repeated Prisoner's Dilemma. We demonstrate that cooperation enforcement is required to achieve the optimal state of resource discovery.

Online social search (OSS) brings forth a new way to harness the Internet for answers. In this paper, we study the non-cooperation problem in OSS. We propose an analytical model that captures the behavior of OSS nodes, and, from a gaming-strategy point of view, analyze various strategies an individual node can utilize to allocate its awareness capacity. Based on this we derive the Pareto inefficiency in terms of the system cost. We also propose an incentive scheme under which the optimal state of individual nodes is also optimal for the whole system. Extensive simulations show that the strategy under our proposed incentive mechanism outperforms other strategies in terms of the system cost and the search success rate. To the best of our knowledge, this is the first study of the tragedy-of-the-commons problem in OSS.

We propose and analyze a multi-server model that captures a performance trade-off between centralized and distributed processing. In our model, a fraction p of an available resource is deployed in a centralized manner (e.g., to serve a most loaded station) while the remaining fraction 1-p is allocated to local servers that can only serve requests addressed specifically to their respective stations.

Most replication-based packet forwarding algorithms in opportunistic networks neglect the fairness issue on the success rate distribution among all participants. In this paper we discuss the fairness evaluation on success rate, and propose a new fair packet forwarding strategy which operates as a plugin for traditional utility-based routing protocols. We compare the performance of our strategy with several well-known routing schemes via both a synthetic contact model and real human mobility traces. We find that our strategy improves the balance of success rates among users while maintaining approximately the same system throughput. In addition, our scheme reduces the cost of traditional utility-based routing protocols.

In this paper, we propose a novel concept for a social TV application targeting the demographic of viewers enjoying live sports events, such as road bicycle racing. We intend to enhance the viewing experiences of spectators with sensor-fitted bikes tied to an interactive biking environment on television. The system enables a new form of personalized, physical, and virtual-reality interaction between viewers and a TV program, as well as interactions within or between communities of friends. We also describe a prototype we have implemented to demonstrate the feasibility of our idea. The prototype, my second bike, uses a 3D mirrored world environment (Google Earth) to visually represent participating spectators, competing athletes and outdoor bikers. We contend that the system has the potential to attract and support a large user base on account of its scalability, ease of deployment and ability to promote audience participation in live sports events on TV.

Online social search brings forth a new way to harness the Internet for answers. However, the personal and often sensitive information is unwittingly exposed to others when a person looks for an expert via the underlying social network. In this paper, we propose a model in which a node's behavior of looking for an expert is adjusted by his awareness of the potential expertise of his contacts. We derive the optimal distribution of nodes' awareness level that minimizes the system's privacy exposure, and prove that it corresponds to the unique Nash equilibrium. Our analysis shows that the optimal distribution over a posed question is inversely proportional to the square root of the corresponding expertise density.

Online social search (OSS) brings forth a new way to harness the Internet for answers. In this paper, we study the balancing between OSS users' capabilities and responsibilities. Targeting a practical system design, we propose an analytical model that captures the heterogeneity of different referral sessions in OSS, and a distributed socio-aware referral strategy that can achieve the desired balance when the system reaches steady state. We show that configuring the strategy enables the system operator to control the flow of all posed questions in the system. We also discuss the implications of configuring the strategy from a gaming-strategy point of view.

In this paper, we derive the general expressions of the throughput capacity scaling of mobile ad hoc networks (MANETs), and find that the strategy of an individual node determines the throughput of an MANET. We show that optimal strategies that maximize the throughput of the network can exist. Based on a game theoretic approach, we further show that the optimal strategies constitute Nash equilibria.

Online social networking systems provide indirect access to a large number of people connected by multi-step chains of acquaintances, and plays an important role in the referrals for human information flow. In this paper, from a networking point of view, we study the problem of locating experts for relevant information via online social networks. We model the action of forwarding a question with random walk, adjusted by a node's awareness of the potential expertise of his immediate neighbours. Using the model we derive analytical expressions of the performance metrics of a referral session in terms of the nodes' awareness level of their neighbours and the percentage of nodes that may have answers to the posed question. We also utilize several real online social networks to study the modeled question-forwarding strategy, and find that the simulation results validate our analyses.

We show that CSMA is able to spontaneously synchronize transmissions in a wireless network with constant-size packets, and that this property can be used to devise efficient synchronized CSMA scheduling mechanisms without message passing. Using tools from queuing theory, we prove that for any connected wireless networks with arbitrary interference constraints, it is possible to implement self-synchronizing TDMA schedules without any explicit message passing or clock synchronization besides transmitting the original data packets, and the interaction can be fully local in that each node decides when to transmit next only by overhearing its neighbors' transmissions. We also provide a necessary and sufficient condition on the emergence of self-synchronization for a given TDMA schedule, and prove that such conditions for self-synchronization can be checked in a finite number of steps for a finite network topology.

In this paper, we propose a hybrid cargo-level tracking system for logistics. We highlight the special system requirements, discuss the design issues and identify the design principles. Then we propose an innovative hybrid system. As far as we know, this is the first system that exploits both infrastructure-based and infrastructure-less positioning schemes for practical cargo-level tracking. Compared with existing systems, the proposed system provides a ubiquitous cargo-level tracking solution with enhanced availability, reliability, and lower total costs.

In opportunistic networks, centrality characterizes a node's capability to act as a communication hub. In this paper, we provide an in-depth study of choosing effective centrality metrics for message forwarding in bandwidth-limited opportunistic networks. Based on this study, we propose a destination-unaware forwarding algorithm that accounts for the popularity of a node and the contact durations between nodes. We evaluate the algorithm on two experimental human mobility traces. The simulation results show that the proposed algorithm achieves higher system throughput while maintaining a lower forwarding cost compared with several known destination-unaware forwarding schemes.

Despite the growing importance of multimedia applications, we still know relatively little about how to specify, design, and maintain this class of complex applications in a systematic manner. The concept of software architecture has recently emerged as a way to improve our ability to effectively construct and maintain large-scale complex software systems. Under this new paradigm, software engineers are able to do evolutionary design of complex systems through architecture specification, design rationale capture, architecture validation and verification, and architecture transformation. Several architecture description languages (ADLs), such as Wright, Rapide, UniCon, ACME, etc. have been proposed to support the architecture development under this new software paradigm. Although current ADLs more or less support certain features of object-oriented design approach, few of them are purely based on object-oriented paradigm. In this paper, we present an architecture description language — OOADL (Object-Oriented Architecture Description Language) to facilitate the architecture specification of multimedia software systems. This language takes object-oriented paradigm as its backbone, and provides formal semantics for modeling architectures of software systems. It also aims at other goals such as, support for hierarchical refinement, support for reuse of architecture styles, support for analysis, and support for exception handling. We also introduce the default architecture style which brings the features of extensibility and re-usability into the language. Finally, we use OOADL to construct part of the architecture framework of a multimedia system to illustrate the usage and modeling capability of OOADL.