Preprints

Journal

~ It was conjectured in 1996 that feedback increases the capacity of queue channels except when the service time distribution is memoryless. We prove the conjecture in the affirmative for bounded service time distributions. A part of this work was presented at ITA Workshop, San Diego, 2023 and at ISIT, Taiwan, 2023.

~ The two publications above are the result of a collaborative project aimed at developing a low-cost, custom IED with edge analytics capability for smart grid; our contribution was in designing and implementing “lightweight” streaming signal compression algorithms and providing theoretical guarantees for their performance. In the 2021 paper, the hardware implementation results of the algorithm are presented. The 2022 paper deals with the theoretical analysis of the proposed algorithm, comparison with the state-of-the-art, and rigorous testing on practical data sets. This is currently the state-of-the-art in practical electrical signal compression for distribution systems.

~ We propose a scheme for analog-to-digital (A/D) conversion of a signal that is bandlimited most of the time but contains harmonics of a high frequency occasionally (as in the case of electrical signals considered above). The proposed scheme performs better in terms of compression compared to the state-of-the-art scheme while retaining the same guarantees on the reconstruction error.

~ We provide shorter proofs for recent results that characterize the p-norm of a certain class of circulant matrices, using Fourier analysis. The key observation is that a circulant matrix is diagonalized by a DFT matrix.

~ We characterize the communication complexity of testing if the correlation between a given d-dimensional vector and a scalar distributed across two nodes, has considerable energy or not, in terms of d.  A part of this work was presented at ISIT, Colorado, 2018.