Jayanth S

I’m an AI/ML Research Specialist at Nokia Standards, currently working on LLM-based retrieval workflows, RAG pipelines, and deep learning models for 6G physical layer research.

I hold an MS (Research) degree from the Indian Institute of Technology, Dharwad, where I worked on the Optimization of Age of Information in Wireless Communication Networks under Asst. Prof. Rajshekhar V Bhat. My research interests span Machine Learning, Reinforcement Learning, Stochastic Optimization, and applied AI systems.

news

Dec 06, 2023	Won the Depth Estimation using mmWave AI/ML challenge - 2023 organised by NIST (National Institute of Standards and Technology, USA) and ITU (International Telecommunication Union). Presentation link

latest posts

May 02, 2022	Equivalence of Forward and Backward view in TD($\lambda$) method
Apr 25, 2022	Monte-Carlo Algorithm with Reward Reshaping for Mountain Car gym environment
Feb 20, 2022	Contraction Property of Bellman Operator with contraction factor $\gamma < 1$

selected publications

Conference

Distortion Minimization with Age of Information and Cost Constraints

S Jayanth, Nikolaos Pappas, and Rajshekhar V Bhat

IEEE WiOpt, Aug 2023

Abs DOI

We consider a source node deployed in a real-time monitoring application that needs to sample a stochastic process and convey its state timely and accurately to a destination over a wireless ON/OFF channel. The source can either process a raw sample to determine its current state and transmit that information or transmit the raw sample and let the destination determine the state. The source is subjected to an average cost constraint, and it cannot sample, process, and transmit at all the time instants due to the associated costs. When the destination does not receive information, it uses the previous information as an estimate of the current state, which, if it matches the actual state at the source, the distortion is considered to be zero. The objective is to minimize average expected distortion subject to constraints on the average expected age of information (AoI) of states of interest and costs incurred by the source, where the AoI of a state increases if no status update is received, else drops to unity. We derive a stationary randomized policy (SRP) to solve the formulated problem, for which we obtain the expression for the expected AoI under the SRP using a lumpability argument on the two-dimensional discrete-time Markov chain formed using AoI and instantaneous distortion as states. We extensively study the impact of the system parameters on the average distortion under the SRP and draw significant conclusions.
Conference

RIBCE: RIS-BS Virtual Array Based Channel Estimation for mm-Wave Communication System

S Jayanth, Vishnu Karthikeya Gorty, A Anil Kumar, and 2 more authors

IEEE ICC Workshops, Jun 2024

Abs DOI

Reconfigurable Intelligent Surface (RIS) is an emerging technology capable of manipulating the propagation environment, expected to play a vital role in all future wireless communication networks. This paper addresses the important and challenging channel estimation (CE) problem of a fully passive RIS-assisted communication system. We combine RIS array elements and Base Station (BS) antenna elements, forming a virtual multi-dimensional array referred to as the RIS-BS virtual array. Leveraging this array, we propose two algorithms: RIBCE-Compressive Sensing (CS) and RIBCE-Estimation of Signal Parameters via Rotational Invariance Techniques (ESPRIT), to jointly estimate the angle of arrival (AoA) and angle of departure (AoD). The corresponding channel gains are then computed using the estimated AoA and AoD. A lower bound on the pilot overhead is also provided with these proposed algorithms which shows a significant reduction in the pilot requirement compared to state-of-art (SoA) approach. Simulation results are also provided which corroborates the theory and shows a significant performance improvement over the state-of-art approach.