Research and Publications

Ali Yekkehkhany

Please see my Google Scholar profile for a complete list of publications.

My research is focused on applied probability theory and stochastic processes with a wide range of applications in online learning and multi-armed bandits, queueing theory, and stochastic game theory. I have also worked extensively on personalized education and e-learning. My research has been recognized by receiving the best poster award in recognition of high-quality research, professional poster, and outstanding presentation in the 15th CSL Student Conference, 2020. Some of my research projects are listed below.

Risk-Averse Online Learning and Multi-Armed Bandits

The term rational has become synonymous with exploiting the arm with maximum expected reward in the context of multi-armed bandits. In contrast, the psychological studies in prospect theory show that human beings are risk-averse in nature. As a result, we advocate the idea of risk aversion in explore-then-commit bandits with the objective of competing against the arm with the best risk-return trade-off. Additionally, we study the trade-off between cost and regret when pulling arms in the exploration phase incurs a cost.

Publications:

A. Yekkehkhany, E. Arian, R. Nagi, and I. Shomorony, ‘‘A Cost-Based Analysis for Risk-Averse Explore-Then-Commit Finite-Time Bandits’’, IISE Transactions, 2021. [link]

A. Yekkehkhany, E. Arian, M. Hajiesmaili, and R. Nagi, ‘‘Risk-Averse Explore-Then-Commit Algorithms for Finite-Time Bandits’’, IEEE Conference on Decision and Control (CDC), 2019. [link]

Risk-Averse (Congestion) Game Theory with Incomplete Information

The term rational has become synonymous with maximizing expected payoff in the definition of the best response in Nash setting. In contrast, we consider games with incomplete information in which players engage for finite number of times. In such games, it may not be rational for players to maximize their expected payoff as they cannot wait for the Law of Large Numbers to take effect. We instead define new notions of risk-averse best responses that result in different classes of risk-averse equilibria for stochastic games. We also study risk aversion in congestion games with incomplete information with applications in smart and automatic navigation systems used in the fast-growing market of autonomous vehicles, unmanned aerial vehicles, and fleets in general.

Publications:

A. Yekkehkhany, T. Murray, and R. Nagi, ‘‘Stochastic Superiority Equilibrium in Game Theory’’, INFORMS, Decision Analysis Journal, 2021. [link]

A. Yekkehkhany, and Rakesh Nagi, ‘‘Risk-Averse Equilibrium for Autonomous Vehicles in Stochastic Congestion Games’’, under submission. [link]

Affinity Load Balancing and Data Centers

The data-intensive analytics in MapReduce, Hadoop, and Dryad frameworks can be modeled by the affinity scheduling problem in queueing theory. The fundamental problem to all these data-parallel applications is data locality. Despite the existence of multiple locality levels within and across data centers, the existing theoretical works analyze systems with only two levels of locality. We find that going from two to three levels of locality changes the problem drastically since a trade-off between performance and throughput emerges. To tackle this issue, we propose the Balanced Priority Algorithm for Near-Data Scheduling (Balanced PANDAS) and establish its throughput optimality. Furthermore, we prove the PANDAS algorithm to be heavy-traffic optimal for up to three locality levels in a specific part of the capacity region. We also propose the Generalized Balanced PANDAS (GB-PANDAS) algorithm that is blind to both task arrival rates and service rates. In such a case, we show through examples that natural learning is not sufficient for stabilizing the system or minimizing the mean task completion time. To tackle this issue, the GB-PANDAS algorithm is equipped with an exploration-exploitation scheme to learn the queueing structure.

Publications:

A. Yekkehkhany, and Rakesh Nagi, ‘‘Blind GB-PANDAS: A blind throughput-optimal load balancing algorithm for affinity scheduling’’, IEEE/ACM Transactions on Networking, 2020. [link]

Q. Xie, A. Yekkehkhany, and Y. Lu, ‘‘Scheduling with multi-level data locality: Throughput and heavy-traffic optimality’’, IEEE INFOCOM, The 35th Annual IEEE International Conference on Computer Communications, 2016. [link]

A. Yekkehkhany, A. Hojjati, and M. Hajiesmaili, ‘‘GB-PANDAS: Throughput and heavy-traffic optimality analysis for affinity scheduling’’, ACM SIGMETRICS Performance Evaluation Review, 2018. [link]

Personalized Education and e-Learning

The traditional teaching methods with fixed curriculum sequences that are widely used at universities for science, technology, engineering, and mathematics (STEM) and K-12 courses do not take different abilities of learners into account. This ignorance of learners’ prior background knowledge, pace of learning, various preferences, and learning goals in the current education system can cause tremendous pain and discouragement for those who do not keep pace with this inefficient system. To put this into perspective, we developed an online personalized education tool, called MasterProbo, aiming at narrowing/closing the achievement gap in STEM and K-12 education, offering a cheaper alternative to private tutoring by recommending personalized curriculum sequences, and benefiting minority groups, including but not limited to women and African-American students in STEM who tend to be more isolated and lose motivation in the current system. We used artificial intelligence techniques, including neural networks and graphical models, to predict the learning proficiency of students and recommend optimal curriculum sequences. MasterProbo has been used by more than 1000 students in 6 semesters for the probability with engineering applications course at the University of Illinois at Urbana-Champaign.

Publications:

D. Su, A. Yekkehkhany, Y. Lu, and W. Lu, ‘‘Prob2Vec: Mathematical Semantic Embedding for Problem Retrieval in Adaptive Tutoring’’, IEEE American Control Conference (ACC), 2020. [link]

Publications

A. Yekkehkhany, and Rakesh Nagi, ‘‘Blind GB-PANDAS: A blind throughput-optimal load balancing algorithm for affinity scheduling’’, IEEE/ACM Transactions on Networking, 2020. [link]

A. Yekkehkhany, T. Murray, and R. Nagi, ‘‘Stochastic Superiority Equilibrium in Game Theory’’, INFORMS, Decision Analysis Journal, 2021. [link]

A. Yekkehkhany, E. Arian, R. Nagi, and I. Shomorony,‘‘A Cost-Based Analysis for Risk-Averse Explore-Then-Commit Finite-Time Bandits’’, IISE Transactions, 2021. [link]

A. Yekkehkhany, and Rakesh Nagi, ‘‘Risk-Averse Equilibrium for Autonomous Vehicles in Stochastic Congestion Games’’, under submission. [link]

Q. Xie, A. Yekkehkhany, and Y. Lu, ‘‘Scheduling with multi-level data locality: Throughput and heavy-traffic optimality’’, IEEE INFOCOM, The 35th Annual IEEE International Conference on Computer Communications, 2016. [link]

A. Yekkehkhany, A. Hojjati, and M. Hajiesmaili, ‘‘GB-PANDAS: Throughput and heavy-traffic optimality analysis for affinity scheduling’’, ACM SIGMETRICS Performance Evaluation Review, 2018. [link]

A. Yekkehkhany, E. Arian, M. Hajiesmaili, and R. Nagi, ‘‘Risk-Averse Explore-Then-Commit Algorithms for Finite-Time Bandits’’, IEEE Conference on Decision and Control (CDC), 2019. [link]

D. Su, A. Yekkehkhany, Y. Lu, and W. Lu, ‘‘Prob2Vec: Mathematical Semantic Embedding for Problem Retrieval in Adaptive Tutoring’’, IEEE American Control Conference (ACC), 2020. [link]

Y. Ruan, A. Yekkehkhany, S. R. Etesami, ‘‘Online Learning for Job Scheduling on Heterogeneous Machines’’, to appear in IEEE Conference on Decision and Control (CDC), 2020.

B. Alinia, M. S. Talebi, M. H. Hajiesmaili, A. Yekkehkhany, and N. Crespi, ‘‘Competitive online scheduling algorithms with applications in deadline-constrained EV charging’’, IEEE/ACM 26th International Symposium on Quality of Service (IWQoS), 2018. [link]

M. Hosseini, Y. Jiang, A. Yekkehkhany, R. R. Berlin, and L. Sha, ‘‘A mobile geo-communication dataset for physiology-aware dash in rural ambulance transport’’, Proceedings of the 8th ACM on Multimedia Systems Conference, 2017. [link]

H. N. Matin, A. Yekkehkhany, and Rakesh Nagi, ‘‘Probabilistic Analysis of UAV Routing with Dynamically Arriving Targets’’, 22th IEEE International Conference on Information Fusion (FUSION), 2019. [link]

A. Ghassami, A. Yekkehkhany, and N. Kiyavash, ‘‘A covert queueing channel in round robin schedulers’’, arXiv preprint arXiv:1701.08883, 2017. [link]

A. Yekkehkhany, ‘‘Near-Data Scheduling for Data Centers with Multiple Levels of Data Locality’’, Master Thesis, University of Illinois at Urbana-Champaign, 2017. [link]

A. Yekkehkhany, ‘‘Risk-Averse Multi-Armed Bandits and Game Theory’’, Ph.D. Dissertation, University of Illinois at Urbana-Champaign, 2020.