I joined IIT (BHU) in 2017. Earlier I completed Ph.D from IISc., Bangalore and had post-doc experience from IIT Bombay, University of Leoben, Austria and National University of Singapore. My research interests are in machine learning where I work mostly in reinforcement learning and its applications. I'm also associated with the Coforge Data and AI Lab and the Indian Knowledge Systems Center of Excellence at IIT (BHU).
EDUCATION
PhD Computer Science and Automation, Indian Institute of Science, Bangalore - 2013
Resource Allocation in Edge Computing: A Game-Theoretic Perspective Sumit Kumar, Ruchir Gupta, K Lakshmanan 2022 IEEE Global Conference on Computing Power and Communication Technologies Globconpt 2022, 2022 Edge computing is a novel computing framework that offers decentralized computing power in order to reduce latency and save bandwidth. This technology is critical in the fast-evolving app market, where app vendors hire edge servers to serve their users. To compete with others, each app vendor wants to employ the edge servers in such a way that it costs the least. Therefore, all app vendors need a solution that minimizes the cost of hiring edge servers. This paper proposes an Edge Server Hiring game (ESHGame), a game-theoretical approach to the problem that formulates it as a potential game. We demonstrate in ESHGame that app vendors collectively achieve at least one pure Nash equilibrium (PNE). We also set an upper bound on the price of stability in this game.
Multicasting: A Game-Theoretic Method for Constructing an Efficient Multicast Tree Sumit Kumar, Ruchir Gupta, K Lakshmanan 4th International Conference on Inventive Research in Computing Applications Icirca 2022 Proceedings, 2022 In multicast routing, the path search process can be formulated as a network game where end-users are considered self-motivated players. Every end-user must pay a cost for the path to get the data from the source. Therefore, end-users want to construct the path from the source at the minimum possible cost. In order to get the minimum cost, end-users will keep modifying paths, which leads to instability. When no end-user can do better by modifying his path, the stable state is considered a Pure Nash equilibrium (PNE). The quality of PNE depends on the cost-sharing mechanism among the end-users of edges. Bringing PNE close to the social optimum (minimum total cost for all end-users) has always been a goal for various cost-sharing strategies. It constructs a cost-effective multicast tree for group communication. Many authors analyzed the cost-sharing mechanisms based on the Shapley value concept, far from the social optimum. Our study aims to identify a weighted cost-sharing mechanism and formulate the path construction process as a Multicast Tree Construction Game (MTCGame). This game employs an MTC algorithm to reach PNE. We prove that at least one PNE always exists in MTCGame. We analyze the quality of the PNE numerically.
A Game-Theoretic Approach for Increasing Resource Utilization in Edge Computing Enabled Internet of Things Sumit Kumar, Ruchir Gupta, K. Lakshmanan, Vipin Maurya IEEE Access, 2022 Edge computing is a new paradigm that reduces latency and saves bandwidth by deploying edge servers in different geographic locations. This technology plays a crucial role in the rapidly growing apps market for IoT devices as app vendors can hire computing resources on edge servers to serve their app users. An effective allocation of edge computing resources to different apps is needed to maximize resource utilization and serve the most app users at the lowest cost. We refer to this as an Edge Resource Allocation (ERA) problem. In this paper, we propose an Edge Resource Allocation Game (ERAGame), a game-theoretic approach that formulates the ERA problem by appropriately pricing the multi-tenant edge servers. The proposed approach gives a Pure Nash Equilibrium (PNE) solution to the ERA problem. For this, we design an ERA algorithm using ERAGame under which the system converges to PNE. For fast convergence to PNE, the edge servers are partitioned into different groups, enabling the ERA algorithm to run in parallel on all edge servers within each group. We prove that ERAGame is a potential game that guarantees at least one PNE under the ERA algorithm. We evaluate that the price of stability of ERAGame is at most O(log n). The performance of the proposed algorithm is examined through simulation.
Transition based discount factor for model free algorithms in reinforcement learning Abhinav Sharma, Ruchir Gupta, K. Lakshmanan, Atul Gupta Symmetry, 2021 Reinforcement Learning (RL) enables an agent to learn control policies for achieving its long-term goals. One key parameter of RL algorithms is a discount factor that scales down future cost in the state’s current value estimate. This study introduces and analyses a transition-based discount factor in two model-free reinforcement learning algorithms: Q-learning and SARSA, and shows their convergence using the theory of stochastic approximation for finite state and action spaces. This causes an asymmetric discounting, favouring some transitions over others, which allows (1) faster convergence than constant discount factor variant of these algorithms, which is demonstrated by experiments on the Taxi domain and MountainCar environments; (2) provides better control over the RL agents to learn risk-averse or risk-taking policy, as demonstrated in a Cliff Walking experiment.
Learning in Function Spaces: An Unified Functional Analytic View of Supervised and Unsupervised Learning K Lakshmanan arXiv preprint arXiv:2603.14272 , 2026 2026
A Short Survey of Averaging Techniques in Stochastic Gradient Methods K Lakshmanan arXiv preprint arXiv:2603.09634 , 2026 2026
Influence maximization independent of seed set size K Lakshmanan Operations Research Letters 62, 107309 , 2025 2025
Recursive Class Field Theory: A Computable Framework for Abelian Extensions K Lakshmanan 2025
On the Minimality of the Conductor in Rank Bounds for Elliptic Curves K Lakshmanan arXiv preprint arXiv:2506.20175 , 2025 2025
Computational complexity of finding subgroups of a given order K Lakshmanan arXiv preprint arXiv:2503.11238 , 2025 2025 Citations: 1
Primes and Bivariate Polynomials without Constant Terms: A Recursive Algorithm K Lakshmanan arXiv preprint arXiv:2405.10519 , 2024 2024
Uncomputability of Global Optima for Nonconvex Functions in the Oracle Model K Lakshmanan arXiv preprint arXiv:2401.09436 , 2023 2023
A deep actor critic reinforcement learning framework for learning to rank V Padhye, K Lakshmanan Neurocomputing 547, 126314 , 2023 2023 Citations: 35
Proximal policy optimization based hybrid recommender systems for large scale recommendations V Padhye, K Lakshmanan, A Chaturvedi Multimedia Tools and Applications 82 (13), 20079-20100 , 2023 2023 Citations: 13
Comment on “Federated learning with differential privacy: Algorithms and performance analysis” K Rajkumar, A Goswami, K Lakshmanan, R Gupta IEEE Transactions on Information Forensics and Security 17, 3922-3924 , 2022 2022 Citations: 12
Resource allocation in edge computing: A game-theoretic perspective S Kumar, R Gupta, K Lakshmanan 2022 IEEE Global Conference on Computing, Power and Communication … , 2022 2022 Citations: 3
Multicasting: A Game-Theoretic Method for Constructing an Efficient Multicast Tree S Kumar, R Gupta, K Lakshmanan 2022 4th International Conference on Inventive Research in Computing … , 2022 2022 Citations: 1
A game-theoretic approach for increasing resource utilization in edge computing enabled internet of things S Kumar, R Gupta, K Lakshmanan, V Maurya IEEE Access 10, 57974-57989 , 2022 2022 Citations: 29
An unsupervised software fault prediction approach using threshold derivation R Kumar, A Chaturvedi, L Kailasam IEEE Transactions on Reliability 71 (2), 911-932 , 2022 2022 Citations: 24
Stochastic arrow-hurwicz algorithm for path selection and rate allocation in self-backhauled mmWave networks A Sharma, K Lakshmanan, R Gupta, A Gupta IEEE Communications Letters 26 (3), 716-720 , 2021 2021 Citations: 2
PILHNB: Popularity, interests, location used hidden Naive Bayesian-based model for link prediction in dynamic social networks AK Singh, K Lakshmanan Neurocomputing 461, 562-576 , 2021 2021 Citations: 25
Link prediction-based influence maximization in online social networks AK Singh, L Kailasam Neurocomputing 453, 151-163 , 2021 2021 Citations: 41
Multi-time scale smoothed functional with nesterov’s acceleration A Sharma, K Lakshmanan, R Gupta, A Gupta IEEE Access 9, 113489-113499 , 2021 2021 Citations: 4
Transition based discount factor for model free algorithms in reinforcement learning A Sharma, R Gupta, K Lakshmanan, A Gupta Symmetry 13 (7), 1197 , 2021 2021 Citations: 7
MOST CITED SCHOLAR PUBLICATIONS
An online actor–critic algorithm with function approximation for constrained markov decision processes S Bhatnagar, K Lakshmanan Journal of Optimization Theory and Applications 153 (3), 688-708 , 2012 2012 Citations: 347
Deep feature learning for histopathological image classification of canine mammary tumors and human breast cancer A Kumar, SK Singh, S Saxena, K Lakshmanan, AK Sangaiah, H Chauhan, ... Information Sciences 508, 405-421 , 2020 2020 Citations: 288
CoMHisP: A novel feature extractor for histopathological image classification based on fuzzy SVM with within-class relative density A Kumar, SK Singh, S Saxena, AK Singh, S Shrivastava, K Lakshmanan, ... IEEE Transactions on Fuzzy Systems 29 (1), 103-117 , 2020 2020 Citations: 52
Improved regret bounds for undiscounted continuous reinforcement learning K Lakshmanan, R Ortner, D Ryabko International conference on machine learning, 524-532 , 2015 2015 Citations: 50
Link prediction-based influence maximization in online social networks AK Singh, L Kailasam Neurocomputing 453, 151-163 , 2021 2021 Citations: 41
A deep actor critic reinforcement learning framework for learning to rank V Padhye, K Lakshmanan Neurocomputing 547, 126314 , 2023 2023 Citations: 35
A game-theoretic approach for increasing resource utilization in edge computing enabled internet of things S Kumar, R Gupta, K Lakshmanan, V Maurya IEEE Access 10, 57974-57989 , 2022 2022 Citations: 29
Beam alignment for mmWave using non-stationary bandits R Gupta, K Lakshmanan, AK Sah IEEE Communications Letters 24 (11), 2619-2622 , 2020 2020 Citations: 28
PILHNB: Popularity, interests, location used hidden Naive Bayesian-based model for link prediction in dynamic social networks AK Singh, K Lakshmanan Neurocomputing 461, 562-576 , 2021 2021 Citations: 25
An unsupervised software fault prediction approach using threshold derivation R Kumar, A Chaturvedi, L Kailasam IEEE Transactions on Reliability 71 (2), 911-932 , 2022 2022 Citations: 24
A novel cloud-assisted secure deep feature classification framework for cancer histopathology images A Kumar, SK Singh, K Lakshmanan, S Saxena, S Shrivastava ACM Transactions on Internet Technology (TOIT) 21 (2), 1-22 , 2021 2021 Citations: 22
Multiscale Q-learning with linear function approximation S Bhatnagar, K Lakshmanan Discrete Event Dynamic Systems 26 (3), 477-509 , 2016 2016 Citations: 15
Proximal policy optimization based hybrid recommender systems for large scale recommendations V Padhye, K Lakshmanan, A Chaturvedi Multimedia Tools and Applications 82 (13), 20079-20100 , 2023 2023 Citations: 13
A novel Q-learning algorithm with function approximation for constrained Markov decision processes K Lakshmanan, S Bhatnagar 2012 50th Annual Allerton Conference on Communication, Control, and … , 2012 2012 Citations: 13
Comment on “Federated learning with differential privacy: Algorithms and performance analysis” K Rajkumar, A Goswami, K Lakshmanan, R Gupta IEEE Transactions on Information Forensics and Security 17, 3922-3924 , 2022 2022 Citations: 12
A parameter-free affinity based clustering B Mukhoty, R Gupta, L K, M Kumar Applied Intelligence 50 (12), 4543-4556 , 2020 2020 Citations: 9
Transition based discount factor for model free algorithms in reinforcement learning A Sharma, R Gupta, K Lakshmanan, A Gupta Symmetry 13 (7), 1197 , 2021 2021 Citations: 7
Quasi-Newton smoothed functional algorithms for unconstrained and constrained simulation optimization K Lakshmanan, S Bhatnagar Computational Optimization and Applications 66 (3), 533-556 , 2017 2017 Citations: 5
Multi-time scale smoothed functional with nesterov’s acceleration A Sharma, K Lakshmanan, R Gupta, A Gupta IEEE Access 9, 113489-113499 , 2021 2021 Citations: 4
Smoothed functional and quasi-Newton algorithms for routing in multi-stage queueing network with constraints K Lakshmanan, S Bhatnagar International Conference on Distributed Computing and Internet Technology … , 2011 2011 Citations: 4
