arXiv subjects

Krishna Jagannathan

Publications and source records attributed to Krishna Jagannathan.

44 records · Page 3Linked to original sources

arXiv2017

Efficient CSMA using Regional Free Energy Approximations

CSMA (Carrier Sense Multiple Access) algorithms based on Gibbs sampling can achieve throughput optimality if certain parameters called the fugacities are appropriately chosen. However, the problem of computing these fugacities is NP-hard. In this work, we derive estimates of the fugacities by using a framework called the regional free energy approximations. In particular, we derive explicit expressions for approximate fugacities corresponding to any feasible service rate vector. We further prove that our approximate fugacities are exact for the class of chordal graphs. A distinguishing feature of our work is that the regional approximations that we propose are tailored to conflict graphs with small cycles, which is a typical characteristic of wireless networks. Numerical results indicate that the fugacities obtained by the proposed method are quite accurate and significantly outperform the existing Bethe approximation based techniques.

cs.NI↗

arXiv2016

Collaborative Learning of Stochastic Bandits over a Social Network

We consider a collaborative online learning paradigm, wherein a group of agents connected through a social network are engaged in playing a stochastic multi-armed bandit game. Each time an agent takes an action, the corresponding reward is instantaneously observed by the agent, as well as its neighbours in the social network. We perform a regret analysis of various policies in this collaborative learning setting. A key finding of this paper is that natural extensions of widely-studied single agent learning policies to the network setting need not perform well in terms of regret. In particular, we identify a class of non-altruistic and individually consistent policies, and argue by deriving regret lower bounds that they are liable to suffer a large regret in the networked setting. We also show that the learning performance can be substantially improved if the agents exploit the structure of the network, and develop a simple learning algorithm based on dominating sets of the network. Specifically, we first consider a star network, which is a common motif in hierarchical social networks, and show analytically that the hub agent can be used as an information sink to expedite learning and improve the overall regret. We also derive networkwide regret bounds for the algorithm applied to general networks. We conduct numerical experiments on a variety of networks to corroborate our analytical results.

cs.LG↗

arXiv2016

Contagion processes on urban bus networks in Indian cities

Bus transportation is considered as one of the most convenient and cheapest modes of public transportation in Indian cities. Due to their cost-effectiveness and wide reachability, they help a significant portion of the human population in cities to reach their destinations every day. Although from a transportation point of view they have numerous advantages over other modes of public transportation, they also pose a serious threat of contagious diseases spreading throughout the city. The presence of numerous local spatial constraints makes the process and extent of epidemic spreading extremely difficult to predict. Also, majority of the studies have focused on the contagion processes on scale-free network topologies whereas, spatially-constrained real-world networks such as, bus networks exhibit a wide-spectrum of network topology. Therefore, we aim in this study to understand this complex dynamical process of epidemic outbreak and information diffusion on the bus networks for six different Indian cities using SI and SIR models. This will allow us to identify epidemic thresholds for these networks which will help us in controlling outbreaks by developing node-based immunization techniques.

physics.soc-ph↗

arXiv2016

Queuing Approaches to Principal-Agent Communication under Information Overload

In the information overload regime, human communication tasks such as responding to email are well-modeled as priority queues, where priority is determined by a mix of intrinsic motivation and extrinsic motivation corresponding to the task's importance to the sender. We view priority queuing from a principal-agent perspective, and characterize the effect of priority-misalignment and information asymmetry between task senders and task receivers in both single-agent and multi-agent settings. In the single-agent setting, we find that discipline can override misalignment. Although variation in human interests leads to performance loss in the single-agent setting, the same variability is useful to the principal with optimal routing of tasks, if the principal has suitable information about agents' priorities. Our approach starts to quantitatively address the effect of human dynamics in routine communication tasks.

cs.IT↗

arXiv2016

Local stability and Hopf bifurcation analysis for Compound TCP

We conduct a local stability and Hopf bifurcation analysis for Compound TCP, with small Drop-tail buffers, in three topologies. The first topology consists of two sets of TCP flows having different round trip times, and feeding into a core router. The second topology corresponds to two queues in tandem, and consists of two distinct sets of TCP flows, regulated by a single edge router and feeding into a core router. The third topology comprises of two distinct sets of TCP flows, regulated by two separate edge routers, and feeding into a common core router. For each of these cases, we conduct a detailed local stability analysis and obtain conditions on the network and protocol parameters to ensure stability. If these conditions get marginally violated, our analysis shows that the underlying systems would lose local stability via a Hopf bifurcation. After exhibiting a Hopf, a key concern is to determine the asymptotic orbital stability of the bifurcating limit cycles. We present a detailed analytical framework to address the stability of the limit cycles, and the type of the Hopf bifurcation by invoking Poincare normal forms and the center manifold theory. We conduct packet-level simulations to highlight the existence and stability of the limit cycles in the queue size dynamics.

math.DS↗

arXiv2015

Spatial CSMA: A Distributed Scheduling Algorithm for the SIR Model with Time-varying Channels

Recent work has shown that adaptive CSMA algorithms can achieve throughput optimality. However, these adaptive CSMA algorithms assume a rather simplistic model for the wireless medium. Specifically, the interference is typically modelled by a conflict graph, and the channels are assumed to be static. In this work, we propose a distributed and adaptive CSMA algorithm under a more realistic signal-to-interference ratio (SIR) based interference model, with time-varying channels. We prove that our algorithm is throughput optimal under this generalized model. Further, we augment our proposed algorithm by using a parallel update technique. Numerical results show that our algorithm outperforms the conflict graph based algorithms, in terms of supportable throughput and the rate of convergence to steady-state.

cs.NI↗

arXiv2014

Finite-Horizon Optimal Transmission Policies for Energy Harvesting Sensors

In this paper, we derive optimal transmission policies for energy harvesting sensors to maximize the utility obtained over a finite horizon. First, we consider a single energy harvesting sensor, with discrete energy arrival process, and a discrete energy consumption policy. Under this model, we show that the optimal finite horizon policy is a threshold policy, and explicitly characterize the thresholds, and the thresholds can be precomputed using a recursion. Next, we address the case of multiple sensors, with only one of them allowed to transmit at any given time to avoid interference, and derive an explicit optimal policy for this scenario as well.

cs.IT↗

arXiv2010

Queue Length Asymptotics for Generalized Max-Weight Scheduling in the presence of Heavy-Tailed Traffic

We investigate the asymptotic behavior of the steady-state queue length distribution under generalized max-weight scheduling in the presence of heavy-tailed traffic. We consider a system consisting of two parallel queues, served by a single server. One of the queues receives heavy-tailed traffic, and the other receives light-tailed traffic. We study the class of throughput optimal max-weight-alpha scheduling policies, and derive an exact asymptotic characterization of the steady-state queue length distributions. In particular, we show that the tail of the light queue distribution is heavier than a power-law curve, whose tail coefficient we obtain explicitly. Our asymptotic characterization also contains an intuitively surprising result - the celebrated max-weight scheduling policy leads to the worst possible tail of the light queue distribution, among all non-idling policies. Motivated by the above negative result regarding the max-weight-alpha policy, we analyze a log-max-weight (LMW) scheduling policy. We show that the LMW policy guarantees an exponentially decaying light queue tail, while still being throughput optimal.

cs.NI↗