Indian Institute of Science, Bangalore Professor, Department of Computer Science and Automation Chairman, Division of Electrical Sciences
Prof. Y. NARAHARI studied in IISc, Bangalore (BE, ME, and PhD) and was an Indo-US Science and Technology Post-Doctoral Fellow in the Massachusetts Institute of Technology, Cambridge, USA, in 1992. He is currently a J.C. Bose National Fellow. His book "Game Theory and Mechanism Design" was recently published by the IISc Press and the World Scientific Publishing Company. Prof. Narahari works in applying game theory and mechanism design to design algorithms that work in the presence of strategic agents. In particular, his current research is focused on mechanism design for multi-armed bandit problems, crowdsourcing, online education, Internet advertising, and blockchain based systems. His recent collaborative R & D projects include: Incentive Design for Enhancing Efficiency and Participation in Online Education (IBM Research); Influence Maximization in the Presence of Strategic Competing Campaigns (Adobe Research Labs, Bangalore); Incentive Compatible Machine Learning (Xerox Corporation) and Intelligent Mechanisms and Algorithms for Carbon Economics (Infosys Technologies, Bangalore). He is an elected Fellow of the IEEE; Indian National Science Academy; Indian Academy of Sciences; Indian National Academy of Engineering; and the National Academy of Sciences
;Blending Mechanism Design with Machine Learning for Solving Emerging Problems in Artificial Intelligence
Mechanism design (MD) provides a game theoretic framework to explore if the given social choice function may be implemented as an equilibrium outcome of an induced game. In a multi-agent setting, machine learning (ML) seeks to learn the preferences or types of the agents through any available data or through intelligent exploration. ML and MD are well investigated as individual problems, however, interesting research questions arise when we try to mix them together. Many emerging problems in AI involve strategic agents holding private information, for example, Internet advertising auctions, Crowdsourcing, smart grids,, smart contracts with blockchains, and online educational platforms. To find a satisfactory solutions to these problems, MD and ML are both required. In this talk, we examine the technically challenging issues that arise when MD and ML have to be blended together. In particular, we focus on our current research into design of multiarmed bandit mechanisms where online learning and mechanism design come together yielding a powerful new modeling option for many important problems in AI.