ANANT SAHAI
Electrical Engineering and Computer Sciences Professor sahai@eecs.berkeley.edu Official website | |

Research For prospective students Bio Teaching My Students |
267 Cory Hall UC Berkeley 94720-1770 Fall 2018 Office Hours: Tu3:30-4:30pm in 258 Cory And by appointment |

My areas of interest span communications, control, and signal processing. Within that range, my focus is on the communications theory side, particularly in the areas of wireless and information theory. Within information theory, my main interest is in developing the conceptual tools needed to understand feedback, interaction, delay, reliability, and complexity. To that end, I am interested in decentralized control systems as they provide well understood mathematical models that do not mesh with the classical notions from information theory. They also tell us why delay and cooperation areimportant. On the wireless communication side, I was interested in power consumption and now am mostly interested in how multi-scale heterogenous wireless systems can coexist peacefully. Cognitive radio and wireless spectrum sharing are of particular interest.

My research sits within the BLISS center, where we study the foundations of information, inference, statistics, systems, and signal processing.

- Decentralized Control
- Built a bridge between control theory and information theory by showing that anytime reliability is
*the right way*to evaluate the quality of a noisy channel being used in a feedback loop for stabilization. Developed tight bounds for anytime reliability. - Obtained the first
*provably*positive results for the longstanding Witsenhausen counterexample in distributed control by demonstrating nonlinear control strategies that are within a constant factor of optimal for all problem parameters. This attack was then extended to give approximately optimal results for the infinite-horizon average-cost two-user decentralized scalar LQG problem. New bounding techniques were developed to do this that generalized cut-set bounds to dynamic programming contexts. - Completely solved the Intermittent Kalman Filtering problem by discovering the concept of eigenvalue cycles and connecting the problem to a characterization of observability rather than to controllability.
- Introduced the concept of control capacity as a fundamental limit for actuation channels in a control loop trying to stabilize a plant.

- Built a bridge between control theory and information theory by showing that anytime reliability is
- Communication and Wireless Networking
- Overturned prior assertions to show how feedback improves the tradeoff between delay and the probability of error. Even noisy/unreliable feedback helps. Moreover, the details of this illuminated an interesting duality between source and channel coding.
- Developed a new bound on how the probability of error must scale with "decoding neighborhood sizes" and used this to illuminate the implementation complexity of codes in terms of power consumption. This showed that for short-range communication, approaching Shannon capacity is probably the wrong thing to aim for.
- Developed a purely algebraic proof of the min-flow=max-cut theorem for network coding. This unified a classical result in decentralized linear control with the seemingly unrelated area of linear network coding.
- Developed a ultra-high-reliability protocol for low-latency communication to support industrial automation and future high-performance Internet-of-Things applications. This synchronized protocol opportunistically leverages multi-user diversity by using distributed space-time codes and network coding and is robust to knowledge of the fading distribution.

- Dynamic Spectrum Sharing
- Showed that modeling uncertainties introduce fundamental limitations (SNR Walls) to spectrum sensing.
- Developed new metrics for evaluating spectrum sensing in the dynamic spectrum access context, introduced the key idea of the spatial overhead of sensing, and showed that diversity (from cooperative sensing) is required to reduce this overhead.
- Introduced a new paradigm for the light-handed regulation of cognitive radios inspired by criminal rather than civil law and explored the resulting regulatory overhead.
- Introduced the idea of white-space maps and developed open-source software for studying spectrum whitespaces and exploring regulatory alternatives using real data in a reproducible way.

Before joining the Berkeley faculty, I spent some time at the startup Enuvis, Inc. where I was on the theoretical/algorithmic side of a team that developed new techniques for GPS detection in very low SNR environments (such as those encountered indoors in urban areas). Before that, I was a graduate student at MIT studying Electrical Engineering and Computer Science (Course 6 in MIT-speak) and was based in the Laboratory for Information and Decision Systems under Prof. Sanjoy Mitter. My research interests there started in machine understanding but shifted toward the intersection of control and information theory. I did my undergraduate work in EECS here at the University of California at Berkeley.

I served as the Treasurer for the IEEE Information Theory Society from 2007-2009 and am currently the faculty adviser for the Berkeley chapter of Eta Kappa Nu.

**EE290S/194**: Machine Learning for Sequential Decision Making under Uncertainty

Term | 16a | 16b | 70 | 120 | 121 | 123 | 126 | 189/289A | 224A | 224B | 226A | 229A | 194/294/290S | 298 | 301 |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|

2018 Fall | X |
||||||||||||||

2018 Spring | X | ||||||||||||||

2017 Fall | X | X | |||||||||||||

2017 Spring | X |
||||||||||||||

2016 Spring | X | ||||||||||||||

2015 Fall | X | ||||||||||||||

2014 Fall | X | ||||||||||||||

2014 Spring | X | ||||||||||||||

2013 Fall | X | ||||||||||||||

2013 Spring | X | ||||||||||||||

2012 Fall | X | ||||||||||||||

2012 Spring | X | ||||||||||||||

2011 Fall | X | X | |||||||||||||

2010 Spring | X | X | |||||||||||||

2009 Fall | X | X | |||||||||||||

2009 Spring | X | X | |||||||||||||

2008 Fall | X | X |
X | ||||||||||||

2008 Spring | X | X | X | ||||||||||||

2007 Fall | X | X | |||||||||||||

2007 Spring | X | X | X | ||||||||||||

2006 Fall | X |
X | |||||||||||||

2006 Spring | X | X | |||||||||||||

2005 Fall | X | X | |||||||||||||

2005 Spring | X | X | |||||||||||||

2004 Fall | X |
X | |||||||||||||

2004 Spring | X | X | |||||||||||||

2003 Fall | X | X | |||||||||||||

2003 Spring | X | X | |||||||||||||

2002 Fall | X | X | |||||||||||||

2002 Spring | X | ||||||||||||||

- Vidya Muthukumar: Working on wireless spectrum-sharing. Helped understand and visualize the impact of the FCC's incentive auction on the TV White Spaces. Currently exploring enforcement issues building on the work of Kristen Ann Woyach as well as information-theoretic fundamental limits on identity.

- Angel Daruna: Spend the summer of 2014 visiting here from Georgia Tech as a part of the SUPERB program. Worked on the FCC incentive auction with Kate Harrison and Vijay Kamble. Represented Berkeley at the Washington DC meeting for REU participants nationwide.
- Hidekazu Miyoshi: visited from SEI Innovation Core in Japan. Explored cognitive radio issues.
- Mukul Agarwal: A student of Prof. Sanjoy Mitter in LIDS. Spent Summer '06 visiting me here in Wireless Foundations. Exploring the nature of information flows in systems.

- Vasuki Narasimha Swamy: Joined the group after doing her MS with Kannan Ramchandran. Internship at Qualcomm Flarion in New Jersey. Developed protocols for ultrareliable low-latency wireless communication networks designed to support high-performance Internet-of-Things devices --- for example, in Industrial automation. Now at Intel Research.
- Kate Harrison: Built on Mubaraq Mishra's work to figure out how to really visualize whitespace opportunities. Built on this in her PhD dissertation on understanding the true potential of the cloud-oriented spectrum management architecture only hinted at in the recent PCAST report on spectrum. Developed a Python-based open-source framework for studying whitespaces. We won best policy paper awards at DySpANs 2012 and 2014. Mentored Vidya Muthukumar and Angel Daruna. Now works for Google in Mountain View.
- Gireeja Ranade: Internships at IBM Research and Microsoft Research. Explored network-extensions of the Witsenhausen counterexample, deterministic models, and the nature of information flow in control systems. She discovered the concept of "control capacity" that shows how actuation uncertainty imposes a fundamental informational bottleneck in our ability to control an unstable system, connecting robust control to information theory. Mentored Govind Ramnarayan, Sahaana Suri, Ena Hariyoshi, and Leah Dickstein as well as Vasuki Narasimha Swami. Co-creator of the EE16AB series along with Elad Alon. Won Angelakos Prize. Now faculty at Berkeley after a few years at Microsoft Research AI in Redmond.
- Kristen Ann Woyach: Great master's thesis on regulatory issues in cognitive radio and spectrum sharing, introducing the concept of spectrum jails. Continued her explorations in this direction for her doctoral dissertation, and also bringing in some spatial stochastic geometry. Co-author on a Leonard G. Abraham Prize winning paper. Mentored Padmini Pyapali. Now works for Google in Pittsburgh.
- Baris Nakiboglu: Postdoc that came to me from MIT, where he had done a wonderful dissertation on topics near to my heart in feedback and information theory. Worked with me on my "white whale" pursuit of understanding for channels with feedback.
- Se Yong Park: Explored the intersection of control theory and information theory for his PhD dissertation. He solved the problem of intermittent Kalman Filtering, connected network coding to decentralized LTI control and thereby showed that implicit communication exists even there, and has greatly built upon the Witsenhausen counterexample to approximately solve the scalar infinite-horizon two-user decentralized LQG control problem. Won the Lim prize for his prelim exam performance and Eliahu Jury prize for his dissertation. Now works for Qualcomm R&D in San Diego.
- Hari Palaiyanur: Master's thesis on applications of sequential decoding to anytime problems. Did summer internship at Nokia Research in '07. In his dissertation, he explored the information theory of active models of perception and how to understand channels with feedback --- discovering a very curious fact about the non-additivity of the classic Haroutunian bound along the way and proving that the sphere-packing bound must hold even for asymmetric DMCs with very delayed feedback. Now works at McKinsey and Company.
- Pulkit Grover: Explored model uncertainty issues from an information theoretic perspective. Summer internship at Wionics Research with Realtek in '07. And PostDoc at Stanford during 2012. Worked with me on understanding the computational complexity of iterative decoding from a power-consumption perspective. Pioneering contributions on the Witsenhausen's counterexample for his doctoral dissertation. We won a best student paper award at IEEE CDC in 2010. Pulkit won the Eliahu Jury award in our department for his dissertation. Co-author on a Leonard G. Abraham Prize winning paper. Now faculty at CMU.
- S. Mubaraq Mishra (Joint with Brodersen in BWRC): Pioneering doctoral research on cooperation and architectural issues for cognitive radios, including the first real analysis of the real-world white-space opportunity. Summer internship at Wionics Research with Realtek in '06. Mentored Kate Harrison. After working for the startup Spidercloud in the Bay Area, he now works for Google Nest.
- Rahul Tandra: Master's thesis on limits to low SNR detection due to uncertainty. His pioneering doctoral thesis built upon this in the context of opportunistic spectrum use and also explored the fundamental metrics/limits for spectrum sensing. We won a best paper award for IEEE DySpAN in 2007. Now working at Qualcomm R&D in San Diego.
- Cheng Chang: Did a Master's thesis on UWB-based sensors: localization and positioning. Doctoral dissertation on multiuser anytime information theory with a focus on multiterminal source coding. Joined HP Labs as postdoc in Jan 08 and is now at the hedge fund D. E. Shaw.
- Danijela Cabric: A student of Bob Brodersen whom I coadvised on spectrum sensing issues. Great thesis and is now tenured faculty at UCLA.
- Niels Hoven: Master's thesis on power control for cognitive radios. Did summer internship at Philips Research in 06. Currently taking an indefinite break from graduate school and is now working at Pocket Gems.
- Stark Draper: Postdoc with myself, Kannan Ramchandran, and David Tse. Worked on feedback and delay issues, especially in the soft deadline context, as well as streaming Slepian Wolf problems. Now tenured faculty at U. Toronto .
- Qing Xu: Master's thesis on anytime capacity of erasure type channels. Joined a Bay Area startup and is now at Qualcomm.
- Tunc Simsek: Brilliant doctoral thesis on anytime coding with feedback. Now at Mathworks. (Joint with Varaiya)

- Leah Dickstein: Worked with Gireeja Ranade at the intersection of control and communication and also with Vasuki Narasimha Swamy on wireless protocols for the high performance Internet-of-Things. Now at Facebook, will join Berkeley for grad school.
- Ena Hariyoshi: Worked with Gireeja Ranade at the intersection of causality, control, and information theory by looking at how valuable non-causal information can be in control settings. Now a graduate student at MIT.
- Sahaana Suri: Worked with Gireeja Ranade on wireless communication protocol for low-latency and high-reliability, especially on the uplink side. Now a graduate student at Stanford.
- Govind Ramnarayan: Worked with Gireeja Ranade on how control over an erasure channel behaves for vector systems with non-causal lookahead. Now a graduate student at MIT.
- Padmini Pyapali: Undergrad researcher with Kristen Woyach where she did some work on enforcing sensing on cognitive radios in a light-handed way. Now at Uber.
- Hansen Bow: Undergrad researcher with me who went on for a PhD at MIT working on BioMEMS. With me, he did some of our earliest work on spectrum sharing as well as work on positioning in UWB systems.
- Moshe Malkin: Undergrad researcher with me who went on to get his PhD at Stanford working with Prof. Cioffi. With me, he did some work on positioning in UWB systems.
- Amirali Zohrenejad: Undergrad researcher with me who is now at Microsoft working in wireless networking. With me, he did some early work on randomized encoding strategies for tracking random walks.

- Profile piece in the MIT LIDS ALL Magazine: mostly featuring the work with Kristen on Spectrum Jails.
- Article in Technology Review, June 2010: focusing on the work that Mubaraq and Kate had done in estimating the value of the TV whitespaces.
- Article in Technology Review, November 2008: featuring my thoughts on the FCC's historic election day ruling.

Special thanks to our past and present research sponsors:

- United States National Science Foundation under grants: ANI-0230963, ANI-0326503, CNS-0403427, CCF-0729122, CCF-0917212, CNS-0932410, CNS-1321155 (project page), ECCS-1343398, AST-1444078, as well as NSF Fellowships for my students.
- Unrestricted Gifts from Futurewei, Sumitomo Electric, and Samsung Corporation
- Vodafone Foundation Fellowships for my students
- C2S2 and others supporting work done in the Berkeley Wireless Research Center.