Journal Articles

Submitted

To appear

1. Spread, then target, and advertise in waves: optimal budget allocation across advertising channels

   By

   S ESHGHI,

   VM Preciado,

   S Sarkar,

   SS Venkatesh,

   Q Zhao,

   R D’Souza,

   A Swami,

   To appear in IEEE Transactions on Network Science and Engineering (TNSE)

   • Abstract
   • BibTeX

   We analyze optimal strategies for the allocation of a finite budget that can be invested in different advertising channels over time with the objective of influencing social opinions in a network of individuals. In our analysis, we consider both exogenous influence mechanisms, such as advertising campaigns, as well as endogenous mechanisms of social influence, such as word-of-mouth and peerpressure, which are modeled using diffusion dynamics. We show that for a general set of objective functions, the optimal influence strategy at every time uses all channels at either their maximum rate or not at all, i.e., a bang-bang strategy. Furthermore, we prove that the number of switches between these extremes is bounded above by a term that is typically much smaller than the number of agents. This means that the optimal influence strategy is to exert maximum effort in waves for every channel, and then cease effort and let the effects propagate. We also show that, at the beginning of the campaign, the total cost-adjusted reach of an exogenous advertising channel determines its relative value. In contrast, as we approach our investment horizon (e.g., election day), the optimal strategy is to invest in channels able to target individuals instead of broad-reaching channels. We demonstrate that the optimal influence strategies are easily computable in several practical cases, and explicitly characterize the optimal controls for the case of linear objective functions in closed form. Finally, we see that, in the canonical example of designing an election campaign, identifying late-deciders is a critical component in the optimal design.

   @article{eshghi2018spread, title = {Spread, then target, and advertise in waves: optimal budget allocation across advertising channels}, author = {ESHGHI, S and Preciado, VM and Sarkar, S and Venkatesh, SS and Zhao, Q and D'Souza, R and Swami, A}, journal = {IEEE Transactions on Network Science and Engineering (TNSE)}, publisher = {IEEE}, status = {accepted}, link = {https://ieeexplore.ieee.org/document/8478819} }

   • Abstract

   We analyze optimal strategies for the allocation of a finite budget that can be invested in different advertising channels over time with the objective of influencing social opinions in a network of individuals. In our analysis, we consider both exogenous influence mechanisms, such as advertising campaigns, as well as endogenous mechanisms of social influence, such as word-of-mouth and peerpressure, which are modeled using diffusion dynamics. We show that for a general set of objective functions, the optimal influence strategy at every time uses all channels at either their maximum rate or not at all, i.e., a bang-bang strategy. Furthermore, we prove that the number of switches between these extremes is bounded above by a term that is typically much smaller than the number of agents. This means that the optimal influence strategy is to exert maximum effort in waves for every channel, and then cease effort and let the effects propagate. We also show that, at the beginning of the campaign, the total cost-adjusted reach of an exogenous advertising channel determines its relative value. In contrast, as we approach our investment horizon (e.g., election day), the optimal strategy is to invest in channels able to target individuals instead of broad-reaching channels. We demonstrate that the optimal influence strategies are easily computable in several practical cases, and explicitly characterize the optimal controls for the case of linear objective functions in closed form. Finally, we see that, in the canonical example of designing an election campaign, identifying late-deciders is a critical component in the optimal design.

   
   

Published

1. Distributed algorithms for multi-layer connected edge dominating sets

   By

   D Papakostas,

   S ESHGHI,

   D Katsaros,

   L Tassiulas,

   IEEE Control Systems Letters (L-CSS)

   vol. 3, no. 1, pp. 31–36, IEEE, 2019

   • Abstract
   • BibTeX

   Monitoring the state of communications in a distributed (multi-layer) network with differing node capabilities requires the maintenance of a backbone which is a connected edge dominating set. In this paper, we present distributed algorithms that provide efficient distributed solutions that can create such multi-layer resilient connected edge-dominating sets. After establishing the complexity of the problem and our proposed heuristics, we experimentally compare their performance while varying multiple characteristics of the underlying networks.

   @article{papakostas2018distributed1, title = {Distributed algorithms for multi-layer connected edge dominating sets }, author = {Papakostas, D and ESHGHI, S and Katsaros, D and Tassiulas, L}, journal = {IEEE Control Systems Letters (L-CSS)}, volume = {3}, number = {1}, pages = {31--36}, year = {2019}, publisher = {IEEE}, link = {https://ieeexplore.ieee.org/document/8398416/} }

   • Abstract

   Monitoring the state of communications in a distributed (multi-layer) network with differing node capabilities requires the maintenance of a backbone which is a connected edge dominating set. In this paper, we present distributed algorithms that provide efficient distributed solutions that can create such multi-layer resilient connected edge-dominating sets. After establishing the complexity of the problem and our proposed heuristics, we experimentally compare their performance while varying multiple characteristics of the underlying networks.

   
   

2. Energy-aware backbone formation in military multilayer ad hoc networks

   By

   D Papakostas,

   S ESHGHI,

   D Katsaros,

   L Tassiulas,

   Ad Hoc Networks

   vol. 81, no. , pp. 17–44, Elsevier, 2018

   • Abstract
   • BibTeX

   Ad hoc networks designed for deployment in modern battlefields need to take care of traditional requirements related to their backbone’s size, their energy efficiency, their scalability in terms of network size, but also of their nature which allows for combining networks of different units that act altogether towards a common operational goal. This article develops a distributed algorithm for developing an energy-efficient backbone for military ad hoc network composed of multiple layers, namely E2CLB. The algorithm is based on the concepts of connected dominating sets and also on node centrality concepts, and results as a heuristic solution to the problem of calculating a maximum energy, minimum connected dominating set for a multilayer network by including into the dominating set those nodes which are highly connected to their and other layers (i.e., they have large centrality value) and moreover they are energy-rich. The computation and communication complexities of the algorithm are analyzed, and a thorough simulation-based evaluation of it against six competitors is presented. The results show that E2CLB is either the best performing algorithm across the examined performance measures or it is able to trade a very small increase in the size of the backbone’s network in order to reap improved performance in the energy realm.

   @article{papakostas2018energy, title = {Energy-aware backbone formation in military multilayer ad hoc networks}, author = {Papakostas, D and ESHGHI, S and Katsaros, D and Tassiulas, L}, journal = {Ad Hoc Networks}, volume = {81}, pages = {17--44}, year = {2018}, publisher = {Elsevier}, link = {https://www.sciencedirect.com/science/article/pii/S1570870518304025} }

   • Abstract

   Ad hoc networks designed for deployment in modern battlefields need to take care of traditional requirements related to their backbone’s size, their energy efficiency, their scalability in terms of network size, but also of their nature which allows for combining networks of different units that act altogether towards a common operational goal. This article develops a distributed algorithm for developing an energy-efficient backbone for military ad hoc network composed of multiple layers, namely E2CLB. The algorithm is based on the concepts of connected dominating sets and also on node centrality concepts, and results as a heuristic solution to the problem of calculating a maximum energy, minimum connected dominating set for a multilayer network by including into the dominating set those nodes which are highly connected to their and other layers (i.e., they have large centrality value) and moreover they are energy-rich. The computation and communication complexities of the algorithm are analyzed, and a thorough simulation-based evaluation of it against six competitors is presented. The results show that E2CLB is either the best performing algorithm across the examined performance measures or it is able to trade a very small increase in the size of the backbone’s network in order to reap improved performance in the energy realm.

   
   

3. Visibility-aware optimal contagion of malware epidemics

   By

   S ESHGHI,

   S Sarkar,

   SS Venkatesh,

   IEEE Transactions on Automatic Control (TAC)

   vol. 62, no. 10, pp. 5205 - 5212, , 2017

   • Abstract
   • BibTeX

   Recent innovations in the design of computer viruses have led to new trade-offs for the attacker. Multiple variants of a malware may spread at different rates and have different levels of visibility to the network. In this work we examine the optimal strategies for the attacker so as to trade off the extent of spread of the malware against the need for stealth. We show that in the mean-field deterministic regime, this spread-stealth trade-off is optimized by computationally simple single-threshold policies. Specifically, we show that only one variant of the malware is spread by the attacker at each time, as there exists a time up to which the attacker prioritizes maximizing the spread of the malware, and after which she prioritizes stealth.

   @article{eshghi2016visibility, author = {ESHGHI, S and Sarkar, S and Venkatesh, SS}, journal = {IEEE Transactions on Automatic Control (TAC)}, title = {Visibility-aware optimal contagion of malware epidemics}, note = {arXiv preprint arXiv:1507.03528}, year = {2017}, volume = {62}, number = {10}, pages = {5205 - 5212}, doi = {10.1109/TAC.2016.2632426}, issn = {0018-9286}, link = {http://ieeexplore.ieee.org/abstract/document/7755729/} }

   • Abstract

   Recent innovations in the design of computer viruses have led to new trade-offs for the attacker. Multiple variants of a malware may spread at different rates and have different levels of visibility to the network. In this work we examine the optimal strategies for the attacker so as to trade off the extent of spread of the malware against the need for stealth. We show that in the mean-field deterministic regime, this spread-stealth trade-off is optimized by computationally simple single-threshold policies. Specifically, we show that only one variant of the malware is spread by the attacker at each time, as there exists a time up to which the attacker prioritizes maximizing the spread of the malware, and after which she prioritizes stealth.

   
   

4. Optimal patching in clustered malware epidemics

   By

   S ESHGHI,

   MHR Khouzani,

   S Sarkar,

   SS Venkatesh,

   IEEE/ACM Transactions on Networking (ToN)

   vol. 24, no. 1, pp. 283–298, IEEE, 2016

   • Abstract
   • BibTeX

   Studies on the propagation of malware in mobile networks have revealed that the spread of malware can be highly inhomogeneous. Platform diversity, contact list utilization by the malware, clustering in the network structure, etc., can also lead to differing spreading rates. In this paper, a general formal framework is proposed for leveraging such heterogeneity to derive optimal patching policies that attain the minimum aggregate cost due to the spread of malware and the surcharge of patching. Using Pontryagin’s Maximum Principle for a stratified epidemic model, it is analytically proven that in the mean-field deterministic regime, optimal patch disseminations are simple single-threshold policies. These policies are amenable to implementation and can serve as benchmarks for policies that have less knowledge of the network. Through numerical simulations, the behavior of optimal patching policies is investigated in sample topologies, and their advantages are demonstrated.

   @article{eshghi2016optimal, title = {Optimal patching in clustered malware epidemics}, author = {ESHGHI, S and Khouzani, MHR and Sarkar, S and Venkatesh, SS}, journal = {IEEE/ACM Transactions on Networking (ToN)}, note = {arXiv preprint arXiv:1403.1639}, volume = {24}, number = {1}, pages = {283--298}, year = {2016}, publisher = {IEEE}, link = {http://ieeexplore.ieee.org/abstract/document/6966800/} }

   • Abstract

   Studies on the propagation of malware in mobile networks have revealed that the spread of malware can be highly inhomogeneous. Platform diversity, contact list utilization by the malware, clustering in the network structure, etc., can also lead to differing spreading rates. In this paper, a general formal framework is proposed for leveraging such heterogeneity to derive optimal patching policies that attain the minimum aggregate cost due to the spread of malware and the surcharge of patching. Using Pontryagin’s Maximum Principle for a stratified epidemic model, it is analytically proven that in the mean-field deterministic regime, optimal patch disseminations are simple single-threshold policies. These policies are amenable to implementation and can serve as benchmarks for policies that have less knowledge of the network. Through numerical simulations, the behavior of optimal patching policies is investigated in sample topologies, and their advantages are demonstrated.

   
   

5. Optimal energy-aware epidemic routing in DTNs

   By

   S ESHGHI,

   MHR Khouzani,

   S Sarkar,

   NB Shroff,

   SS Venkatesh,

   IEEE Transactions on Automatic Control (TAC)

   vol. 60, no. 6, pp. 1554–1569, IEEE, 2015

   • Abstract
   • BibTeX

   In this work, we investigate the use of epidemic routing in energy constrained delay tolerant networks (DTNs). In epidemic routing, messages are relayed by intermediate nodes at contact opportunities, i.e., when pairs of nodes come within the transmission range of each other. Each node needs to decide whether to forward its message upon contact with a new node based on its own residual energy level and the age of that message. We mathematically characterize the fundamental trade-off between energy conservation and a measure of Quality of Service as a dynamic energy-dependent optimal control problem. We prove that in the mean-field regime, the optimal dynamic forwarding decisions follow simple threshold-based structures in which the forwarding threshold for each node depends on its current remaining energy. We then characterize the nature of this dependence. Our simulations reveal that the optimal dynamic policy significantly outperforms heuristics.

   @article{eshghi2015optimal, title = {Optimal energy-aware epidemic routing in DTNs}, author = {ESHGHI, S and Khouzani, MHR and Sarkar, S and Shroff, NB and Venkatesh, SS}, journal = {IEEE Transactions on Automatic Control (TAC)}, volume = {60}, number = {6}, pages = {1554--1569}, note = {arXiv preprint arXiv:1403.1642}, year = {2015}, publisher = {IEEE}, link = {http://ieeexplore.ieee.org/abstract/document/7027154/} }

   • Abstract

   In this work, we investigate the use of epidemic routing in energy constrained delay tolerant networks (DTNs). In epidemic routing, messages are relayed by intermediate nodes at contact opportunities, i.e., when pairs of nodes come within the transmission range of each other. Each node needs to decide whether to forward its message upon contact with a new node based on its own residual energy level and the age of that message. We mathematically characterize the fundamental trade-off between energy conservation and a measure of Quality of Service as a dynamic energy-dependent optimal control problem. We prove that in the mean-field regime, the optimal dynamic forwarding decisions follow simple threshold-based structures in which the forwarding threshold for each node depends on its current remaining energy. We then characterize the nature of this dependence. Our simulations reveal that the optimal dynamic policy significantly outperforms heuristics.

   
   

Conferences and Workshops

To appear

1. Energy-aware distributed edge domination of multilayer networks

   By

   D Papakostas,

   S ESHGHI,

   D Katsaros,

   L Tassiulas,

   To be presented at 2019 American Control Conference (ACC)

   2019

   • Abstract
   • BibTeX

   Monitoring and maintaining communications in a multilayer ad hoc wireless network requires a stable communication overlay. Such monitoring should be resilient, avoiding reliance on a single (or a few) layers, and energy-aware, not being vulnerable to the exhaustion of available energy in a single network element. Thus, there is possibly a trade-off between overlay size, interlayer connectivity and the energy distribution within the overlay. In this paper, we present three distributed energy-aware multilayer connected edge dominating set algorithms, show how they manage such trade-offs in practical scenarios, and show that CCEDS, a pruned centrality-based distributed algorithm, has the best performance.

   @inproceedings{papakostas2018energz, title = {Energy-aware distributed edge domination of multilayer networks}, author = {Papakostas, D and ESHGHI, S and Katsaros, D and Tassiulas, L}, booktitle = {2019 American Control Conference (ACC)}, status = {accepted}, year = {2019} }

   • Abstract

   Monitoring and maintaining communications in a multilayer ad hoc wireless network requires a stable communication overlay. Such monitoring should be resilient, avoiding reliance on a single (or a few) layers, and energy-aware, not being vulnerable to the exhaustion of available energy in a single network element. Thus, there is possibly a trade-off between overlay size, interlayer connectivity and the energy distribution within the overlay. In this paper, we present three distributed energy-aware multilayer connected edge dominating set algorithms, show how they manage such trade-offs in practical scenarios, and show that CCEDS, a pruned centrality-based distributed algorithm, has the best performance.

   
   

Published

1. Efficient Influence Maximization Under Network Uncertainty

   By

   S ESHGHI,

   S Maghsudi,

   V Restocchi,

   S Stein,

   L Tassiulas,

   In INFOCOM Workshop on the Communications and Networking Aspects of Online Social Networks (CAOS)

   2019

   • Abstract
   • BibTeX

   We consider the influence maximization (IM) problem in a partially visible social network. The goal is to design a decision-making framework for an autonomous agent to select a limited set of influential seed nodes to spread a message as widely as possible across the network. We consider the realistic case where only a partial section of the network is visible to the agent, while the rest is one of a finite set of known structures, each with a given realization probability. We show that solving the IM problem in this setting is NP-hard, and we provide analytical guarantees for the performance of a novel computationally-efficient seed-selection approximation algorithm for the agent. In empirical experiments on real-world social networks, we demonstrate the efficiency of our scheme and show that it outperforms state-of-the-art approaches that do not model the uncertainty.

   @inproceedings{eshghi2019efficient, title = {Efficient Influence Maximization Under Network Uncertainty}, author = {ESHGHI, S and Maghsudi, S and Restocchi, V and Stein, S and Tassiulas, L}, booktitle = {INFOCOM Workshop on the Communications and Networking Aspects of Online Social Networks (CAOS)}, year = {2019}, link = {https://www.research.ed.ac.uk/portal/files/82194312/Efficient_Influence_Maximization_ESHGI_DoA220219_AFV.pdf} }

   • Abstract

   We consider the influence maximization (IM) problem in a partially visible social network. The goal is to design a decision-making framework for an autonomous agent to select a limited set of influential seed nodes to spread a message as widely as possible across the network. We consider the realistic case where only a partial section of the network is visible to the agent, while the rest is one of a finite set of known structures, each with a given realization probability. We show that solving the IM problem in this setting is NP-hard, and we provide analytical guarantees for the performance of a novel computationally-efficient seed-selection approximation algorithm for the agent. In empirical experiments on real-world social networks, we demonstrate the efficiency of our scheme and show that it outperforms state-of-the-art approaches that do not model the uncertainty.

   
   

2. Whistleblowing games in networks

   By

   S ESHGHI,

   L Tassiulas,

   In 57th IEEE Conference on Decision and Control (CDC)

   2018

   • Abstract
   • BibTeX

   We model the effect of networks on the uptake of risky innovations by workers and the policy-making of their employers as a Stackelberg game. We individuals can report specific instances of observed risky behavior to management, in contrast with anonymous cases where only the existence of such behavior is reported without mention of potential culprits. We contrast the subgame-perfect equilibria of the semi-anonymous model for general networks and for regular graphs with the corresponding anonymous whistleblowing case, in which whistleblowing has been shown to only flourish in a light-punishment regime. We observe that semi-anonymous whistleblowing can lead to better equilibrium outcomes for the manager, as the network structure induces differences in policies chosen by the workers which can be exploited by the manager to maintain moderate amounts of risky innovation. Workers will choose degreedependent threshold policies that instruct them to cheat if the audit probability is below the threshold and to report observed cheating otherwise. Without this network-induced heterogeneity (e.g., for regular graphs), the outcome of semi-anonymous whistleblowing resembles that of anonymous policies. Finally, we show that in a light-punishment regime, workers with the fewest neighbors will be the likeliest to cheat.

   @inproceedings{eshghi2018whistleblowing, title = {Whistleblowing games in networks}, author = {ESHGHI, S and Tassiulas, L}, booktitle = {57th IEEE Conference on Decision and Control (CDC)}, year = {2018}, link = {https://ieeexplore.ieee.org/abstract/document/8619552} }

   • Abstract

   We model the effect of networks on the uptake of risky innovations by workers and the policy-making of their employers as a Stackelberg game. We individuals can report specific instances of observed risky behavior to management, in contrast with anonymous cases where only the existence of such behavior is reported without mention of potential culprits. We contrast the subgame-perfect equilibria of the semi-anonymous model for general networks and for regular graphs with the corresponding anonymous whistleblowing case, in which whistleblowing has been shown to only flourish in a light-punishment regime. We observe that semi-anonymous whistleblowing can lead to better equilibrium outcomes for the manager, as the network structure induces differences in policies chosen by the workers which can be exploited by the manager to maintain moderate amounts of risky innovation. Workers will choose degreedependent threshold policies that instruct them to cheat if the audit probability is below the threshold and to report observed cheating otherwise. Without this network-induced heterogeneity (e.g., for regular graphs), the outcome of semi-anonymous whistleblowing resembles that of anonymous policies. Finally, we show that in a light-punishment regime, workers with the fewest neighbors will be the likeliest to cheat.

   
   

3. Distributed algorithms for multi-layer connected edge-dominating sets

   By

   D Papakostas,

   S ESHGHI,

   D Katsaros,

   L Tassiulas,

   In 57th IEEE Conference on Decision and Control (CDC)

   2018

   • Abstract
   • BibTeX

   Monitoring the state of communications in a distributed (multi-layer) network with differing node capabilities requires the maintenance of a backbone which is a connected edge dominating set. In this paper, we present distributed algorithms that provide efficient distributed solutions that can create such multi-layer resilient connected edge-dominating sets. After establishing the complexity of the problem and our proposed heuristics, we experimentally compare their performance while varying multiple characteristics of the underlying networks.

   @inproceedings{papakostas2018distributed, title = {Distributed algorithms for multi-layer connected edge-dominating sets }, author = {Papakostas, D and ESHGHI, S and Katsaros, D and Tassiulas, L}, booktitle = {57th IEEE Conference on Decision and Control (CDC)}, year = {2018}, link = {https://ieeexplore.ieee.org/document/8398416/} }

   • Abstract

   Monitoring the state of communications in a distributed (multi-layer) network with differing node capabilities requires the maintenance of a backbone which is a connected edge dominating set. In this paper, we present distributed algorithms that provide efficient distributed solutions that can create such multi-layer resilient connected edge-dominating sets. After establishing the complexity of the problem and our proposed heuristics, we experimentally compare their performance while varying multiple characteristics of the underlying networks.

   
   

4. Innovation, cheating, and whistleblowing - a game theoretic perspective

   By

   S ESHGHI,

   L Tassiulas,

   In 52nd Annual Conference on Information Sciences and Systems (CISS)

   2018

   • Abstract
   • BibTeX

   We model the coupled decision-making of workers and the policy-making of an organization in response to an effort-saving, yet risky innovation. We examine the cases where individuals can only alert the organization of wide-spread wrong-doing (anonymous whistleblowing) and where they can report specific illicit behavior (semi-anonymous reporting). We show that depending on policy parameters and individual characteristics, one of 4 equilibria may arise, with the interesting outcomes that whistleblowing only flourishes in a light-punishment regime, and that semi-anonymous reporting leads to the relative prevalence of illicit behavior.

   @inproceedings{eshghi2018innovation, title = {Innovation, cheating, and whistleblowing - a game theoretic perspective}, author = {ESHGHI, S and Tassiulas, L}, booktitle = {52nd Annual Conference on Information Sciences and Systems (CISS)}, year = {2018}, link = {https://ieeexplore.ieee.org/abstract/document/8362286} }

   • Abstract

   We model the coupled decision-making of workers and the policy-making of an organization in response to an effort-saving, yet risky innovation. We examine the cases where individuals can only alert the organization of wide-spread wrong-doing (anonymous whistleblowing) and where they can report specific illicit behavior (semi-anonymous reporting). We show that depending on policy parameters and individual characteristics, one of 4 equilibria may arise, with the interesting outcomes that whistleblowing only flourishes in a light-punishment regime, and that semi-anonymous reporting leads to the relative prevalence of illicit behavior.

   
   

5. A computational framework for modelling inter-group behaviour using psychological theory

   By

   RE Bellamy,

   GB Colombo,

   S ESHGHI,

   G De Mel,

   C Giammanco,

   R Morris,

   DG Rand,

   L Turner,

   RM Whitaker,

   GR Williams,

   In SPIE Defense + Commercial Sensing 2018

   2018

   • Abstract
   • BibTeX

   Psychological theories of inter-group behaviour oer justied representations for interaction, influence, and motivation for coalescence. Agent-based modelling of this behaviour, using evolutionary approaches, further provides a powerful tool to examine the implications of these theories in a dynamic context. In particular, this can enhance our understanding of the escalation of hostility and warfare, and its mitigation, contributing to policy and interventions. In this paper we propose a framework through which social psychology can be embedded in computation for the examination of inter-group behaviour. We examine how various social-psychological theories can be embedded in evolutionary models, and identify ways in which visualisation can support the objective assessment of emergent behaviour. We also discuss how real-world data can be used to parameterise scenarios on which modelling is conducted.

   @inproceedings{bellamy2018computational, title = {A computational framework for modelling inter-group behaviour using psychological theory}, author = {Bellamy, RE and Colombo, GB and ESHGHI, S and De Mel, G and Giammanco, C and Morris, R and Rand, DG and and Turner, L and Whitaker, RM and Williams, GR}, booktitle = {SPIE Defense + Commercial Sensing 2018}, year = {2018}, link = {https://www.spiedigitallibrary.org/conference-proceedings-of-spie/10653/106530G/A-computational-framework-for-modelling-inter-group-behaviour-using-psychological/10.1117/12.2309621.full?SSO=1} }

   • Abstract

   Psychological theories of inter-group behaviour oer justied representations for interaction, influence, and motivation for coalescence. Agent-based modelling of this behaviour, using evolutionary approaches, further provides a powerful tool to examine the implications of these theories in a dynamic context. In particular, this can enhance our understanding of the escalation of hostility and warfare, and its mitigation, contributing to policy and interventions. In this paper we propose a framework through which social psychology can be embedded in computation for the examination of inter-group behaviour. We examine how various social-psychological theories can be embedded in evolutionary models, and identify ways in which visualisation can support the objective assessment of emergent behaviour. We also discuss how real-world data can be used to parameterise scenarios on which modelling is conducted.

   
   

6. Spread, then target, and advertise in waves: optimal capital allocation across advertising channels

   By

   S ESHGHI,

   VM Preciado,

   S Sarkar,

   SS Venkatesh,

   Q Zhao,

   R D’Souza,

   A Swami,

   In IEEE Information Theory and Applications Workshop (ITA)

   2017

   • Abstract
   • BibTeX

   We obtain optimal strategies for the allocation of influence budget across multiple channels and across time for an external influencer, e.g., a political campaign, seeking to maximize its effect on an election given a network of agents with linear consensus-seeking opinion dynamics. We show that for a general set of objective functions, the optimal influence strategy at every time uses all channels at either their maximum rate or not at all. Furthermore, we prove that the number of switches between these extremes is bounded above both by a term that is typically much smaller than the number of agents. This means that the optimal influence strategy is to exert maximum effort in waves for every channel, and then cease effort and let the effects propagate. We also show that at the beginning, the total cost-adjusted reach of a channel determines its relative value, while targeting matters more closer to election time. We demonstrate that the optimal influence structures are easily computable in several practical cases. We explicitly characterize the optimal controls for the case of linear objective functions via a closed form. Finally, we see that in the canonical election example, identifying late-deciders approximately determines the optimal campaign resource allocation strategy.

   @inproceedings{eshghi2017spread, title = {Spread, then target, and advertise in waves: optimal capital allocation across advertising channels}, author = {ESHGHI, S and Preciado, VM and Sarkar, S and Venkatesh, SS and Zhao, Q and D'Souza, R and Swami, A}, booktitle = {IEEE Information Theory and Applications Workshop (ITA)}, year = {2017}, link = {http://ita.ucsd.edu/workshop/17/files/paper/paper_220.pdf} }

   • Abstract

   We obtain optimal strategies for the allocation of influence budget across multiple channels and across time for an external influencer, e.g., a political campaign, seeking to maximize its effect on an election given a network of agents with linear consensus-seeking opinion dynamics. We show that for a general set of objective functions, the optimal influence strategy at every time uses all channels at either their maximum rate or not at all. Furthermore, we prove that the number of switches between these extremes is bounded above both by a term that is typically much smaller than the number of agents. This means that the optimal influence strategy is to exert maximum effort in waves for every channel, and then cease effort and let the effects propagate. We also show that at the beginning, the total cost-adjusted reach of a channel determines its relative value, while targeting matters more closer to election time. We demonstrate that the optimal influence structures are easily computable in several practical cases. We explicitly characterize the optimal controls for the case of linear objective functions via a closed form. Finally, we see that in the canonical election example, identifying late-deciders approximately determines the optimal campaign resource allocation strategy.

   
   

7. Efficient dynamic centrality metrics for election advertising - a case study

   By

   S ESHGHI,

   L Tassiulas,

   In Yale Day of Data 2017 (Poster)

   2017

   • Abstract
   • BibTeX

   In prior work, we have shown how optimal choice of advertising channels for a budget-constrained electoral campaign can be structured. In this poster, we apply the resulting proposed algorithm to the MIT Social Evolution data-set (N=84), which captured political discussions, inclinations, and voting behaviors around the 2008 US Presidential Election within a student dorm. We compare the resulting centrality metrics developed from our algorithm (which have a direct mapping to optimal channel choice decisions) against more traditional static centralities.

   @inproceedings{eshghi2017efficient, title = {Efficient dynamic centrality metrics for election advertising - a case study}, author = {ESHGHI, S and Tassiulas, L}, booktitle = {Yale Day of Data 2017 (Poster)}, year = {2017}, link = {https://elischolar.library.yale.edu/dayofdata/2017/posters/} }

   • Abstract

   In prior work, we have shown how optimal choice of advertising channels for a budget-constrained electoral campaign can be structured. In this poster, we apply the resulting proposed algorithm to the MIT Social Evolution data-set (N=84), which captured political discussions, inclinations, and voting behaviors around the 2008 US Presidential Election within a student dorm. We compare the resulting centrality metrics developed from our algorithm (which have a direct mapping to optimal channel choice decisions) against more traditional static centralities.

   
   

8. Influence maximisation in partially observable social networks

   By

   S Stein,

   S ESHGHI,

   S Maghsudi,

   L Tassiulas,

   RE Bellamy,

   NR Jennings,

   In Workshop on Distributed Analytics InfraStructure and Algorithms for Multi-Organization Federations (DAIS)

   2017

   • Abstract
   • BibTeX

   We consider the problem of choosing influential members within a social network, in order to disseminate a message as widely as possible. While this so-called problem of influence maximisation has been widely studied, little work considers partially-observable networks, where only part of a network is visible to the decision maker. Yet, this is critical in many applications, where an organisation needs to distribute its message far beyond its boundaries and beyond its usual sphere of influence. In this paper, we show that existing algorithms are not sufficient to handle such scenarios. To address this, we propose a set of novel adaptive algorithms that perform well in partially observable settings, achieving an up to 18% improvement on the non-adaptive state of the art.

   @inproceedings{stein2017influence, title = {Influence maximisation in partially observable social networks}, author = {Stein, S and ESHGHI, S and Maghsudi, S and Tassiulas, L and Bellamy, RE and Jennings, NR}, booktitle = {Workshop on Distributed Analytics InfraStructure and Algorithms for Multi-Organization Federations (DAIS)}, year = {2017}, link = {https://dais-ita.org/dais2017} }

   • Abstract

   We consider the problem of choosing influential members within a social network, in order to disseminate a message as widely as possible. While this so-called problem of influence maximisation has been widely studied, little work considers partially-observable networks, where only part of a network is visible to the decision maker. Yet, this is critical in many applications, where an organisation needs to distribute its message far beyond its boundaries and beyond its usual sphere of influence. In this paper, we show that existing algorithms are not sufficient to handle such scenarios. To address this, we propose a set of novel adaptive algorithms that perform well in partially observable settings, achieving an up to 18% improvement on the non-adaptive state of the art.

   
   

9. A framework for modelling the effect of emotion on uncritical reasoning

   By

   D Mott,

   T Kelley,

   C Giammanco,

   S ESHGHI,

   Y Zhang,

   In Workshop on Knowledge Systems for Coalition Operations (KSCO)

   2017

   • Abstract
   • BibTeX

   We describe research on understanding group mutability in the behaviour of external groups, and how interventions by coalition forces may affect the behaviour in terms of controlling hostile groups and encouraging friendly groups. We explore how emotion may influence the behaviour of individuals by affecting the type of reasoning that they undertake, encouraging “uncritical” rather than “critical” thinking. We describe a computational framework holding a cognitive model of an individual operating within a group context, inspired by theories from social science. Individuals relate to in-groups and out-groups and have beliefs that are associated with emotions. Cognitive Appraisal Theory is used to evaluate incoming memes “pronounced” by external speakers, appraising the effects of the memes on an individual’s self-esteem taking account of their group relationships as indicated by social identity theory, and leading to an emotion in the individual. Appraisal is followed by a process of coping that seeks to handle the effects by either performing problem-focused (critical) or emotion-focused (uncritical) thinking, according to the current emotional state of the individual. This model is implemented within a Cognitive Architecture (Soar) as a set of reasoning processes that handle beliefs and emotion. The model is integrated into a multi-agent simulation tool (Repast Simphony) allowing the simulation of populations of individuals interacting and spreading rumours, or memes, together with interventions. We describe how this framework could be used to construct experiments to explore how different situations lead to group mutability and behaviour, together with the effects of interventions by coalition forces.

   @inproceedings{mott2017framework, title = {A framework for modelling the effect of emotion on uncritical reasoning}, author = {Mott, D and Kelley, T and Giammanco, C and ESHGHI, S and Zhang, Y}, booktitle = {Workshop on Knowledge Systems for Coalition Operations (KSCO)}, year = {2017}, link = {http://www.aiai.ed.ac.uk/project/coalition/ksco-2017.html} }

   • Abstract

   We describe research on understanding group mutability in the behaviour of external groups, and how interventions by coalition forces may affect the behaviour in terms of controlling hostile groups and encouraging friendly groups. We explore how emotion may influence the behaviour of individuals by affecting the type of reasoning that they undertake, encouraging “uncritical” rather than “critical” thinking. We describe a computational framework holding a cognitive model of an individual operating within a group context, inspired by theories from social science. Individuals relate to in-groups and out-groups and have beliefs that are associated with emotions. Cognitive Appraisal Theory is used to evaluate incoming memes “pronounced” by external speakers, appraising the effects of the memes on an individual’s self-esteem taking account of their group relationships as indicated by social identity theory, and leading to an emotion in the individual. Appraisal is followed by a process of coping that seeks to handle the effects by either performing problem-focused (critical) or emotion-focused (uncritical) thinking, according to the current emotional state of the individual. This model is implemented within a Cognitive Architecture (Soar) as a set of reasoning processes that handle beliefs and emotion. The model is integrated into a multi-agent simulation tool (Repast Simphony) allowing the simulation of populations of individuals interacting and spreading rumours, or memes, together with interventions. We describe how this framework could be used to construct experiments to explore how different situations lead to group mutability and behaviour, together with the effects of interventions by coalition forces.

   
   

10. Heuristic algorithms for influence maximization in partially observable social networks

    By

    S ESHGHI,

    S Maghsudi,

    V Restocchi,

    S Stein,

    L Tassiulas,

    RE Bellamy,

    NR Jennings,

    In International Workshop on Social Influence Analysis (SocInf)

    2017

    • Abstract
    • BibTeX

    We consider the problem of selecting the most influential members within a social network, in order to disseminate a message as widely as possible. This problem, also referred to as \textitseed selection for influence maximization, has been under intensive investigation since the emergence of social networks. Nonetheless, a large body of existing research is based on the assumption that the network is completely known, whereas little work considers partially observable networks. Yet, due to many issues including the extremely large size of current networks and privacy considerations, assuming full knowledge of the network is rather unrealistic. Despite this, an influencer often wishes to distribute its message far beyond the boundaries of the known network. In this paper, we propose a set of novel heuristic algorithms that specifically target nodes at this boundary, in order to maximize influence across the whole network. We show that these algorithms outperform the state of the art by up to 38% in networks with partial observability.

    @inproceedings{eshghi2017heuristic, title = {Heuristic algorithms for influence maximization in partially observable social networks}, author = {ESHGHI, S and Maghsudi, S and Restocchi, V and Stein, S and Tassiulas, L and Bellamy, RE and Jennings, NR}, booktitle = {International Workshop on Social Influence Analysis (SocInf)}, year = {2017}, link = {http://socinf2017.isistan.unicen.edu.ar/} }

    • Abstract

    We consider the problem of selecting the most influential members within a social network, in order to disseminate a message as widely as possible. This problem, also referred to as \textitseed selection for influence maximization, has been under intensive investigation since the emergence of social networks. Nonetheless, a large body of existing research is based on the assumption that the network is completely known, whereas little work considers partially observable networks. Yet, due to many issues including the extremely large size of current networks and privacy considerations, assuming full knowledge of the network is rather unrealistic. Despite this, an influencer often wishes to distribute its message far beyond the boundaries of the known network. In this paper, we propose a set of novel heuristic algorithms that specifically target nodes at this boundary, in order to maximize influence across the whole network. We show that these algorithms outperform the state of the art by up to 38% in networks with partial observability.

    
    

11. Social group stability and fracture

    By

    S ESHGHI,

    GR Williams,

    GB Colombo,

    L Turner,

    DG Rand,

    RM Whitaker,

    L Tassiulas,

    In 55th Annual Allerton Conference on Communication, Control, and Computing (Allerton)

    2017

    • Abstract
    • BibTeX

    In this paper, we present a mathematical model for the mutation of social groups. Group mutability has been studied in multiple domains, with insights generated on significant factors at differing scales. Mathematical modeling enables the simultaneous study of such phenomena, understanding interactions and generating hypotheses for experiments. In particular, we focus on group fracture, where individuals leave groups of which they are members. For example, this can be due to perceived differences with other group members due to norm related conflict (such as extreme actions by some members). Our aim is to consider simple mathematical models incorporating a selection of social and psychological theory which describes these phenomena as a way to understand their interplay, and describe the trade-offs and challenges.

    @inproceedings{eshghi2017social, title = {Social group stability and fracture}, author = {ESHGHI, S and Williams, GR and Colombo, GB and Turner, L and Rand, DG and Whitaker, RM and Tassiulas, L}, booktitle = {55th Annual Allerton Conference on Communication, Control, and Computing (Allerton)}, year = {2017}, link = {http://ieeexplore.ieee.org/document/8262776/} }

    • Abstract

    In this paper, we present a mathematical model for the mutation of social groups. Group mutability has been studied in multiple domains, with insights generated on significant factors at differing scales. Mathematical modeling enables the simultaneous study of such phenomena, understanding interactions and generating hypotheses for experiments. In particular, we focus on group fracture, where individuals leave groups of which they are members. For example, this can be due to perceived differences with other group members due to norm related conflict (such as extreme actions by some members). Our aim is to consider simple mathematical models incorporating a selection of social and psychological theory which describes these phenomena as a way to understand their interplay, and describe the trade-offs and challenges.

    
    

12. Mathematical models for social group behavior

    By

    S ESHGHI,

    GR Williams,

    GB Colombo,

    L Turner,

    DG Rand,

    RM Whitaker,

    L Tassiulas,

    In Workshop on Distributed Analytics InfraStructure and Algorithms for Multi-Organization Federations (DAIS)

    2017

    • Abstract
    • BibTeX

    In this paper, we seek to identity how mathematical and economic analysis can be used to gain insights about the mutation of social groups. Group mutability has been studied in multiple domains, with insights generated on significant factors at differing scales. Mathematical modeling enables the simultaneous study of such phenomena, understanding interactions and generating hypotheses for experiments. In particular, we focus on group fracture, where individuals leave groups of which they are members. For example, this can be due to perceived differences with other group members due to norm related conflict (such as extreme actions by some members). Our aim is to consider simple mathematical models incorporating a selection of social and psychological theory which describes these phenomena as a way to understand their interplay, and describe the trade-offs and challenges. This will help a federation model the behavior of extremist groups, and determine not only when an intervention is necessary, but also the best course of action to take to induce the fracture of such groups. This paper is an exploratory investigation into methods of achieving this goal and evaluating the usefulness of the outputs to federations.

    @inproceedings{eshghi2017mathematical, title = {Mathematical models for social group behavior}, author = {ESHGHI, S and Williams, GR and Colombo, GB and Turner, L and Rand, DG and Whitaker, RM and Tassiulas, L}, booktitle = {Workshop on Distributed Analytics InfraStructure and Algorithms for Multi-Organization Federations (DAIS)}, year = {2017}, link = {https://dais-ita.org/dais2017} }

    • Abstract

    In this paper, we seek to identity how mathematical and economic analysis can be used to gain insights about the mutation of social groups. Group mutability has been studied in multiple domains, with insights generated on significant factors at differing scales. Mathematical modeling enables the simultaneous study of such phenomena, understanding interactions and generating hypotheses for experiments. In particular, we focus on group fracture, where individuals leave groups of which they are members. For example, this can be due to perceived differences with other group members due to norm related conflict (such as extreme actions by some members). Our aim is to consider simple mathematical models incorporating a selection of social and psychological theory which describes these phenomena as a way to understand their interplay, and describe the trade-offs and challenges. This will help a federation model the behavior of extremist groups, and determine not only when an intervention is necessary, but also the best course of action to take to induce the fracture of such groups. This paper is an exploratory investigation into methods of achieving this goal and evaluating the usefulness of the outputs to federations.

    
    

13. Heuristic algorithms for influence maximization in partially observable social networks

    By

    S Stein,

    S ESHGHI,

    S Maghsudi,

    L Tassiulas,

    RE Bellamy,

    NR Jennings,

    In 7th International Conference on Complex Networks and Their Applications (COMPLEX NETWORKS)

    2017

    • Abstract
    • BibTeX

    We consider the problem of selecting the most influential members within a social network, in order to disseminate a message as widely as possible. This problem, also referred to as \textitseed selection for influence maximization, has been under intensive investigation since the emergence of social networks. Nonetheless, a large body of existing research is based on the assumption that the network is completely known, whereas little work considers partially observable networks. Yet, due to many issues including the extremely large size of current networks and privacy considerations, assuming full knowledge of the network is rather unrealistic. Despite this, an influencer often wishes to distribute its message far beyond the boundaries of the known network. In this paper, we propose a set of novel heuristic algorithms that specifically target nodes at this boundary, in order to maximize influence across the whole network. We show that these algorithms outperform the state of the art by up to 38% in networks with partial observability.

    @inproceedings{stein2017heuristic, title = {Heuristic algorithms for influence maximization in partially observable social networks}, author = {Stein, S and ESHGHI, S and Maghsudi, S and Tassiulas, L and Bellamy, RE and Jennings, NR}, booktitle = {7th International Conference on Complex Networks and Their Applications (COMPLEX NETWORKS)}, year = {2017}, link = {http://complexnetworks.org/index2017.html} }

    • Abstract

    We consider the problem of selecting the most influential members within a social network, in order to disseminate a message as widely as possible. This problem, also referred to as \textitseed selection for influence maximization, has been under intensive investigation since the emergence of social networks. Nonetheless, a large body of existing research is based on the assumption that the network is completely known, whereas little work considers partially observable networks. Yet, due to many issues including the extremely large size of current networks and privacy considerations, assuming full knowledge of the network is rather unrealistic. Despite this, an influencer often wishes to distribute its message far beyond the boundaries of the known network. In this paper, we propose a set of novel heuristic algorithms that specifically target nodes at this boundary, in order to maximize influence across the whole network. We show that these algorithms outperform the state of the art by up to 38% in networks with partial observability.

    
    

14. Visibility-aware optimal contagion of malware epidemics

    By

    S ESHGHI,

    S Sarkar,

    SS Venkatesh,

    In IEEE Information Theory and Applications Workshop (ITA)

    2015

    • BibTeX

    @inproceedings{eshghi2015optimal1, title = {Visibility-aware optimal contagion of malware epidemics}, author = {ESHGHI, S and Sarkar, S and Venkatesh, SS}, booktitle = {IEEE Information Theory and Applications Workshop (ITA)}, year = {2015}, link = {http://ita.ucsd.edu/workshop/15/talks} }
    
    

15. Optimal battery pricing and energy management for microgrids

    By

    S ESHGHI,

    RM Patil,

    In 2015 American Control Conference (ACC)

    2015

    • Abstract
    • BibTeX

    This paper develops an optimal dispatch strategy for a microgrid’s resources through battery pricing. A microgrid fulfills local demand using local generation (conventional and renewable) and storage, along with the electric grid (macrogrid). Demand and renewables such as Photovoltaic (PV) generation are uncertain throughout the day. Optimal management of the resources in a microgrid is necessary to integrate the uncertain renewable energy, serve the uncertain demand and provide the highest returns in terms of cost or emissions. The approach developed in this paper requires minimal modeling effort with regards to uncertainties and emphasizes battery pricing using realized profiles of demand and PV generation. We show that the optimal decision at each time instant is an economic dispatch that takes a rigorously-defined shadow-price for the battery power as input. Furthermore, we show that the calculation of this shadow-price can be simplified to avoid state-constraints in the computation of the optimal control solution. Finally, the results highlight the practical utility of this approach and the developed quasi-optimal solution is comparable to the deterministic optimum cost.

    @inproceedings{eshghi2015optimal2, title = {Optimal battery pricing and energy management for microgrids}, author = {ESHGHI, S and Patil, RM}, booktitle = {2015 American Control Conference (ACC)}, pages = {4994--5001}, year = {2015}, organization = {IEEE}, link = {http://ieeexplore.ieee.org/abstract/document/7172117/} }

    • Abstract

    This paper develops an optimal dispatch strategy for a microgrid’s resources through battery pricing. A microgrid fulfills local demand using local generation (conventional and renewable) and storage, along with the electric grid (macrogrid). Demand and renewables such as Photovoltaic (PV) generation are uncertain throughout the day. Optimal management of the resources in a microgrid is necessary to integrate the uncertain renewable energy, serve the uncertain demand and provide the highest returns in terms of cost or emissions. The approach developed in this paper requires minimal modeling effort with regards to uncertainties and emphasizes battery pricing using realized profiles of demand and PV generation. We show that the optimal decision at each time instant is an economic dispatch that takes a rigorously-defined shadow-price for the battery power as input. Furthermore, we show that the calculation of this shadow-price can be simplified to avoid state-constraints in the computation of the optimal control solution. Finally, the results highlight the practical utility of this approach and the developed quasi-optimal solution is comparable to the deterministic optimum cost.

    
    

16. Optimal patching in clustered epidemics of malware

    By

    MHR Khouzani,

    S ESHGHI,

    S Sarkar,

    SS Venkatesh,

    In IEEE Information Theory and Applications Workshop (ITA)

    2012

    • BibTeX

    @inproceedings{eshghi2012optimal, title = {Optimal patching in clustered epidemics of malware}, author = {Khouzani, MHR and ESHGHI, S and Sarkar, S and Venkatesh, SS}, booktitle = {IEEE Information Theory and Applications Workshop (ITA)}, year = {2012}, link = {http://ita.ucsd.edu/workshop/12/talks} }
    
    

17. Optimal energy-aware epidemic routing in DTNs

    By

    MHR Khouzani,

    S ESHGHI,

    S Sarkar,

    NB Shroff,

    SS Venkatesh,

    In 13th ACM international symposium on Mobile Ad Hoc Networking and Computing (MobiHoc)

    2012

    • Abstract
    • BibTeX

    In this work, we investigate the use of epidemic routing in energy constrained Delay Tolerant Networks (DTNs). In DTNs, connected paths between source and destination rarely materialize due to the mobility and sparse density of nodes. Epidemic routing is well-suited for these environments due to its simplicity and fully distributed implementation. In epidemic routing, messages are relayed by intermediate nodes at contact opportunities, i.e., when pairs of nodes come within transmission range. Each node needs to decide whether to forward its message upon contact with a new node based on its residual energy level and the age of that message. We mathematically characterize the fundamental trade-off between energy conservation and forwarding efficacy as a heterogeneous dynamic energy-dependent optimal control problem. We prove, somewhat surprisingly given the complex nature of the problem, that in the mean field regime, the optimal dynamic forwarding decisions follow simple threshold-based structures in which the forwarding threshold for each node depends on its current remaining energy. We analytically establish this result under generalized classes of utility functions for DTNs. We then characterize the dependence of these thresholds on current energy reserves in each node.

    @inproceedings{khouzani2012optimal, title = {Optimal energy-aware epidemic routing in DTNs}, author = {Khouzani, MHR and ESHGHI, S and Sarkar, S and Shroff, NB and Venkatesh, SS}, booktitle = {13th ACM international symposium on Mobile Ad Hoc Networking and Computing (MobiHoc)}, pages = {175--182}, year = {2012}, organization = {ACM}, link = {http://dl.acm.org/citation.cfm?id=2248397} }

    • Abstract

    In this work, we investigate the use of epidemic routing in energy constrained Delay Tolerant Networks (DTNs). In DTNs, connected paths between source and destination rarely materialize due to the mobility and sparse density of nodes. Epidemic routing is well-suited for these environments due to its simplicity and fully distributed implementation. In epidemic routing, messages are relayed by intermediate nodes at contact opportunities, i.e., when pairs of nodes come within transmission range. Each node needs to decide whether to forward its message upon contact with a new node based on its residual energy level and the age of that message. We mathematically characterize the fundamental trade-off between energy conservation and forwarding efficacy as a heterogeneous dynamic energy-dependent optimal control problem. We prove, somewhat surprisingly given the complex nature of the problem, that in the mean field regime, the optimal dynamic forwarding decisions follow simple threshold-based structures in which the forwarding threshold for each node depends on its current remaining energy. We analytically establish this result under generalized classes of utility functions for DTNs. We then characterize the dependence of these thresholds on current energy reserves in each node.

    
    

Thesis

1. Optimal control of epidemics in the presence of heterogeneity

   By

   S ESHGHI,

   University of Pennsylvania

   2015

   • Abstract
   • BibTeX

   We seek to identify and address how different types of heterogeneity affect the optimal control of epidemic processes in social, biological, and computer networks. Epidemic processes encompass a variety of models of propagation that are based on contact between agents. Assumptions of homogeneity of communication rates, resources, and epidemics themselves in prior literature gloss over the heterogeneities inherent to such networks and lead to the design of sub-optimal control policies. However, the added complexity that comes with a more nuanced view of such networks complicates the generalizing of most prior work and necessitates the use of new analytical methods. We first create a taxonomy of heterogeneity in the spread of epidemics. We then model the evolution of heterogeneous epidemics in the realms of biology and sociology, as well as those arising from practice in the fields of communication networks (e.g., DTN message routing) and security (e.g., malware spread and patching). In each case, we obtain computational frameworks using Pontryagin’s Maximum Principle that will lead to the derivation of dynamic controls that optimize general, context-specific objectives. We then prove structures for each of these vectors of optimal controls that can simplify the derivation, storage, and implementation of optimal policies. Finally, using simulations and real-world traces, we examine the benefits achieved by including heterogeneity in the control decision, as well as the sensitivity of the models and the controls to model parameters in each case.

   @phdthesis{eshghi2015optimal3, title = {Optimal control of epidemics in the presence of heterogeneity}, author = {ESHGHI, S}, school = {University of Pennsylvania}, year = {2015}, link = {http://search.proquest.com/docview/1759630563} }

   • Abstract

   We seek to identify and address how different types of heterogeneity affect the optimal control of epidemic processes in social, biological, and computer networks. Epidemic processes encompass a variety of models of propagation that are based on contact between agents. Assumptions of homogeneity of communication rates, resources, and epidemics themselves in prior literature gloss over the heterogeneities inherent to such networks and lead to the design of sub-optimal control policies. However, the added complexity that comes with a more nuanced view of such networks complicates the generalizing of most prior work and necessitates the use of new analytical methods. We first create a taxonomy of heterogeneity in the spread of epidemics. We then model the evolution of heterogeneous epidemics in the realms of biology and sociology, as well as those arising from practice in the fields of communication networks (e.g., DTN message routing) and security (e.g., malware spread and patching). In each case, we obtain computational frameworks using Pontryagin’s Maximum Principle that will lead to the derivation of dynamic controls that optimize general, context-specific objectives. We then prove structures for each of these vectors of optimal controls that can simplify the derivation, storage, and implementation of optimal policies. Finally, using simulations and real-world traces, we examine the benefits achieved by including heterogeneity in the control decision, as well as the sensitivity of the models and the controls to model parameters in each case.

   
   

Patents

Pending

1. Optimal battery pricing and energy management for localized energy resources

   By

   S ESHGHI,

   RM Patil,

   R Sharma,

   US Patent 20,160,093,002

   2016

   • Abstract
   • BibTeX

   A method and system are provided for managing local energy resources of an energy system. The method includes determining, offline by a processor, an offline optimal resource allocation of the local energy resources using a Pontryagin Maximum Principle that involves a continuous, unbounded state. The method further includes determining, in real-time by the processor, a real-time optimal resource allocation of the local energy resources using offline-determined energy storage shadow pricing. The method also includes managing, by the processor, an allocation of the local energy resources in accordance with the offline optimal resource allocation and the real-time optimal resource allocation.

   @misc{patil2016optimal, title = {Optimal battery pricing and energy management for localized energy resources}, author = {ESHGHI, S and Patil, RM and Sharma, R}, year = {2016}, month = mar, journal = {US Patent 20,160,093,002}, link = {https://www.google.com/patents/US20160093002} }

   • Abstract

   A method and system are provided for managing local energy resources of an energy system. The method includes determining, offline by a processor, an offline optimal resource allocation of the local energy resources using a Pontryagin Maximum Principle that involves a continuous, unbounded state. The method further includes determining, in real-time by the processor, a real-time optimal resource allocation of the local energy resources using offline-determined energy storage shadow pricing. The method also includes managing, by the processor, an allocation of the local energy resources in accordance with the offline optimal resource allocation and the real-time optimal resource allocation.

   
   

Invited Talks

1. Social Influence

   1. Dynamic control of spreading processes on networks
      1. Yale University, Yale Institute for Network Science Colloquium, Yale Institute for Network Science (YINS), January 2019
   2. Decision-making tools for influence propagation in social systems
      1. University of Michigan, Communication and Signal Processing Seminar, Electrical Engineering and Computer Science Department, July 2018
   3. Social influence maximization: a synthesis
      1. Yale University, Human Nature Lab (HNL), Yale Institute for Network Science (YINS), March 2018
   4. Influence in social systems
      1. Yale University, Yale Institute for Network Science Seminar, Yale Institute for Network Science (YINS), July 2017

2. Whistleblowing

   1. You scratch my back, and (maybe) I'll scratch yours: whistleblowing games on networks
      1. Yale University, Yale Institute for Network Science Seminar, Yale Institute for Network Science (YINS), June 2018

3. Epidemic Control

   1. Optimal control of epidemics (and opinions) in the presence of heterogeneity
   1. Cornell University, Electrical and Computer Engineering Department, July 2015 (remotely)
   2. Georgetown University, Bansal Lab, Biology Department, July 2016
   3. Yale University, Yale Institute for Network Science Seminar, Yale Institute for Network Science (YINS), July 2016
   4. University of Pennsylvania, Complex Systems Group, Electrical and Systems Engineering Department, July 2016
   5. University of Georgia, Rohani Lab, Biology Department, July 2016
   6. Pennsylvania State University, Ferrari Lab, Center for Infectious Disease Dynamics (CIDD), Biology Department, July 2016 (remotely)
   7. Harvard University, Center for Communicable Disease Dynamics (CCDD), Harvard T.H. Chan School of Public Health, July 2016
   8. Yale University, Crawford Lab, Yale School of Public Health, July 2018