## Talks## Coming UpTitle: Reputation-Based Information Design for Reducing Energy Consumption Date: November 14, 2018 Abstract: Electricity utility providers would like to reduce the total power consumed by its low-tension residential consumer segment. Supply to this segment is often subsidised, and the saved power can be diverted to more profitable segments. Alternatively, the provider may be keen on earning carbon credits by inducing reduced consumption in this segment. The societal network in which the consumers reside may have a prevailing norm such as saving power is environment-friendly and is considered good. Those that consume less are considered more prosocial and may derive a larger reputational benefit. How can the service provider, who is familiar with the prevailing norm and the consumptions of all users, design suitable feedback signals that exploit reputation benefits to reduce global consumption? We call this a problem of information design and address this question in this paper. In this paper, we consider a continuum of agents, wherein each agent has a different intrinsic motivation to reduce her power consumption. Each agent models the power consumption of the others via a distribution. Using this distribution, they will anticipate their reputational benefit and choose a power consumption by trading off their own intrinsic motivation to do a prosocial action, the cost of this prosocial action and their reputation. We assume that the service provider can provide three noisy feedbacks of the power consumption. We will study which one is the best. For each feedback, we are able to characterize a Mean Field Equilibrium, using a fixed point equation. For two specific feedbacks, we prove the uniqueness of the Mean Field Equilibrium. For the last one, we are able to prove that only two Mean Field Equilibria can exist. Also, we prove that one of the feedbacks will always result in lesser global power consumption than the other. Finally, we numerically study the sensitivity of the different parameters over the Mean Field Equilibrium. Besides validating our mathematical results, we are interested in drawing a map of the impact of the different parameters over the Mean Field Equilibrium. The results of this study are not restricted to the framework of energy efficiency but also to congestion problems or resource sharing problems. Speaker Bio: Alexandre Reiffers is a post-doctoral fellow at Robert Bosch Centre for Cyber-Physical Systems. He received the B.Sc. degree in mathematics (2010) from the University of Marseille, the master degree in applied mathematics (2012) from the University of Pierre et Marie CURIE and the Ph.D. degree in computer science (January 2015) from the INRIA (National research institute in computer science and control) and the University of Avignon. His supervisors were Eitan Altman and Yezekael Hayel. From July 2016 to December 2017, Alexandre Reiffers was a researcher at SafranTech where he was working on comparison of maintenance strategies. Most of his research projects concern the application of mathematical tools (game theory, optimization, stochastic process and machine learning) for a better understanding of real-world problems. The different issues that he studies touch topics such as social networks, speech between human and computer, economy and manufacturing. ## PreviousTitle: Recovery of distributed quantum information from a quantum erasure Date: October 3, 2018 Abstract: Entanglement is a key quantum phenomenon which distributes quantum information among many qubits. We use graphs to represent the network information in the form of nodes and edges namely G=(V,E). We use quantum networks to store quantum information in a distributed setting by having a qubit present at each node. We study the scenario of loss of quantum information ensuing due to the failure of a node. By modeling the failure of a node as a quantum erasure, we propose recovery methods to restore the stored quantum information motivated by the Schmidt decomposition and purification. Speaker Bio: Ankur received M.E. degree in Telecommunication from ECE Dept., IISc in 2013 and B.Tech. in ECE from NIT Kurukshetra in 2010. He worked for Ericsson Global Services Pvt. Ltd. Noida between June 2010 and July 2011. Title: Exciting excitons in layered materials Date: Sep 5, 2018 Abstract: Since the discovery of graphene, the two-dimensional materials have attracted a lot of attention among the research community. These two-dimensional materials have strong in-plane bonding and weak out-of-plane bonding from van der Waals forces. The class of transitional metal dichalcogenides (TMDCs) among layered materials is an intriguing platform to explore optoelectronic device design because of their strong light-matter coupling. A pair of electron and hole created by light illumination are strongly bound by coulomb attraction in these materials by virtue of quantum confinement and reduced screening. Commonly known as excitons, these bound e-h pairs give rise to many interesting phenomena like valley sensitive polarization of photocarriers. Excitons are hosted in energy states below the conduction band which determine the optical bandgap of a material. Thus, it is important to understand and engineer these exciton states for various device applications. Band structure of excitons is calculated using Bethe-Salpeter equation framework where electron-hole interaction is incorporated in the free particle band structure Hamiltonian. A complete analysis of exciton band structure helps us to understand bound e-h pair distribution in real and momentum space along with their dynamics. Hamiltonian varies according to the number of layers and thus exciton distribution across a few layer 2D material itself deals with a lot of interesting physical processes. Depending of the real space distribution, there are intra and inter-layer excitons. These excitons in and across the layers exhibit different levels of coulomb interaction and thus decay at different rates. Further, the exciton states can be shifted by applying field where there can be conversion of intralayer to interlayer excitons. The interplay of interlayer and intralayer excitons can be explored to design new optoelectronic devices Speaker Bio: Sarthak got his B.Tech degree from KGEC in 2014 and his M.Tech degree from Jadavpur University in 2016. He joined Ph.D. in 2016 at ECE with Dr. Kausik Majumdar. His research focuses on exciton bandstructure of 2D materials. Title: Rate-Optimal Streaming Codes for Channels with Burst and Isolated Erasures Date: Abstract: Recovery of data packets from packet erasures in a timely manner is critical for many streaming applications. An early paper by Martinian and Sundberg introduced a framework for streaming codes and designed rate-optimal codes that permit delay-constrained recovery from an erasure burst of length up to B. A recent work by Badr et al. extended this result and introduced a channel model that accounts for both burst and isolated erasures. Furthermore, they obtained a rate upper bound for streaming codes that can recover with a time delay T, from any erasure patterns permissible under this generalized model. However, constructions matching the bound were absent, except for a few parameter sets. In this work, we present a family of codes that achieves the rate upper bound for all feasible parameters. Speaker Bio: Nikhil Krishnan M. received the B.Tech. degree in electronics and communication engineering from the Amrita School of Engineering, Kollam, in 2011, and the M.E. degree in telecommunication from the Department of ECE, Indian Institute of Science (IISc), Bengaluru, in 2013. He is currently a Ph.D. student in the same department, working with Prof. P. Vijay Kumar. His research interests include coding theory and information theory, with applications to distributed storage systems. Title: Computational approaches for Structural Analysis of Indian Art Music signals Date: July 27, 2018 Abstract: The implications of internet age for digital music consumption is seen in terms of easy access to large collections of music in recorded formats as well as seamless access to online/live music content. This necessitates culturally-aware content based analysis of music signals for a variety of applications such as music archival, organization, segmentation, retrieval and recommendation applications, automatic music transcription and music synthesis. We have proposed computational approaches to model melodic and timbral aspects of performances. In this seminar, we address modeling high-level aspects of melodic contours of music performances from acoustic features. Raga, the melodic framework of Indian Art Music, is challenging to model owing to its complex grammatical structures, ornamentations and improvisations. In order to compare structural similarities among ragas, stochastic models are used to analyze note patterns present in prescriptive notations. This can be limiting due to various factors such as lack of ornamentation related information in prescriptive notations and non-availability of notations for impromptu forms of renditions. Hence, we embark on melodic contour based data-driven approaches. We address analysis of structural similarities among renditions of ragas. We first estimate the tonic frequency of the melodic contour using a stochastic model. The critical points of the melodic contour are quantized on to the melodic-temporal grid to obtain perceptually acceptable descriptive transcription at note level while retaining the raga characteristics. Repetitive note patterns of variable and fixed lengths are derived using stochastic models. We propose a latent variable approach for raga distinction based on these patterns. The posterior probability of the latent variable is shown to capture similarities across raga renditions. We show that it is possible to visualize the similarities in a low-dimensional embedded space. Experiments show that it is possible to compare and contrast relations and distances between ragas in the embedded space with the musicological knowledge of the same for both Hindustani and Carnatic music forms. The proposed approach also shows robustness to duration of rendition. Speaker Bio: Ranjani is a PhD student in the Department of ECE, IISc, working with Prof. T. V. Sreenivas. Her research interests include music and speech signal processing. Title: Sparse support recovery via covariance estimation Date:
Abstract: In this talk, we will look at the problem of recovering the common support of a set of k-sparse vectors from compressive measurements and its connection to the problem of covariance estimation. Specifically, we have L vectors of dimension N with the same (unknown) support S of size k, and for each vector we observe a noisy version of its projection onto an m-dimensional subspace of R^N. The goal is to recover S from these compressive measurements. We will consider a Bayesian setting where we impose a Gaussian prior with mean zero and diagonal covariance on the unknown vectors, and formulate the support recovery problem as one of covariance estimation. We will see that the maximum likelihood estimate for the covariance matrix can be obtained as the solution to a non negative quadratic program. Using this approach one can recover the support even when k>m (with L large enough), which is not possible using conventional support recovery algorithms. Speaker Bio: Lekshmi Ramesh is a PhD student in the Department of ECE, IISc, working with Prof. Chandra R. Murthy and Prof. Himanshu Tyagi. Her research interests include sparse signal recovery and estimation theory. Title: Extra Samples can Reduce the Communication for Independence Testing Date: 13 June 2018 Abstract: Two parties observing sequences of bits want to determine if their bits were generated independently or not. To that end, the first party communicates to the second. A simple communication scheme involves taking as few sample bits as determined by the sample complexity of independence testing and sending it to the second party. But is there a scheme that uses fewer bits of communication than the sample complexity, perhaps by observing more sample bits? We show that the answer to this question is in the affirmative when the joint distribution is a binary symmetric source. More generally, for any given joint distribution, we present a distributed independence test that uses linear correlation between functions of the observed random variables. Furthermore, we provide lower bounds for the generalisetting that use hypercontractivity and reverse hypercontractivity to obtain a measure change bound between the joint and the independent distributions. The resulting bounds are tight for both a binary symmetric source and a Gaussian symmetric source. Speaker Bio: K.R. Sahasranand is a PhD student in the Department of ECE, working with Dr. Himanshu Tyagi. His research interests include information theory, detection and estimation theory and distributed statistical inference. Title: Optimal Lossless Source codes for Timely Updates Date: 13 June 2018
Abstract: A transmitter observing a sequence of independent and identically distributed random variables seeks to keep a receiver updated about its latest observations. The receiver need not be apprised about each symbol seen by the transmitter, but needs to output a symbol at each time instant t. If at time t the receiver outputs the symbol seen by the transmitter at time U(t) ≤ t, the age of information at the receiver at time t is t − U(t). We study the design of lossless source codes that enable transmission with minimum average age at the receiver. We show that the asymptotic minimum average age can be attained (up to a constant bits gap) by Shannon codes for a tilted version of the original pmf generating the symbols, which can be computed easily by solving an optimization problem. Underlying our construction for minimum average age codes is a new variational formula for integer moments of random variables, which may be of independent interest. Speaker Bio: Prathamesh is currently pursuing Ph.D. in the Department of ECE at IISc under the guidance of Prof. Himanshu Tyagi. Previously, he worked for TCS. He has a Master's from Industrial Engineering and Operations Research, IIT Bombay in 2015 and, a Bachelors in Electronics and Communication Engineering from K.J.Somaiya College of Engineering, Mumbai in 2013. His research interests lie in the areas of Information Theory, Distributed Optimization, Applied Probability. Currently, He is working on designing communication protocols for Timely updates and, Distributed Optimization. Title: First Order Induced Current Imaging and Electrical Properties Tomography Date: May 9, 2018 (Wednesday) Slides The method manages to reconstruct conductivity profiles of in-vivo measurements without the boundary artefacts found in more commonly used Helmholtz-based EPT methods. It is also inherently more robust to noise because only first-order differencing is required as opposed to second-order differencing as in Helmholtz-based approaches. Moreover, reconstructions can be performed in less than a second, allowing for essentially real-time electrical property mapping. The approach presented here provides a novel look at B1 based electrical properties mapping combining the speed of differencing based approaches with the robustness of the integral maxwell based approaches to provide a practical approach for in-vivo applications. Bio: Patrick Fuchs is a PhD student at the Delft University of Technology currently working in a collaboration with K.V.S Hari at the IISc on low power MRI, speeding up MRI scans and electromagnetic modelling of MRI systems. |