all by date all by type workshops courses colloquium seminars lunches
ECE/ITA Colloquium
Friday, 11.16.12, 11 AM - 12 PM, Jacobs Hall, Room 2512 (Booker Conference Room)
What Risks Lead to Ruin?
Venkat Anantharam, UC Berkeley
Insurance transfers losses from the insuree to the insurer for a price per unit time, the premium. The pace of modern technology throws up scenarios where it is difficult to have confidence about what the loss distribution is. For instance, how would one insure potential losses incurred by entities operating on the Internet? Nevertheless, insurance has many benefits: by aggregating risk, it allows for more risk-taking by innovators...more >
Thursday, 11.15.12, 3:00 PM - 4:00 PM, Jacobs Hall, Room 2512 (Booker Conference Room)
Empirical Processes, Typical Sequences and Coordinated Actions
Maxim Raginsky, UIUC
Recent work by Cuff, Permuter and Cover on coordination via communication has revealed a fascinating aspect of strongly typical sequences --- they can be thought of as a means of reproducing relative frequencies of symbols emitted by stationary memoryless sources using finite communication resources. This viewpoint has a number of useful and far-reaching implications in such settings as distributed control and decision-making, network security, and statistical learning...more >
Friday, 11.09.12, 10 a.m  - 11 a.m, Jacobs Hall, Room 2512 (Booker Conference Room)
The Genetic Codes Governing Gene Regulation
Brendan Frey, University of Toronto
In September, the international ENCODE consortium reported that 80% of human genome sequence is used to control the expression of genes, which comprise only 1% of the genome. However, it is not yet clear how those elements combine to form the logic circuits that dictate gene expression in response to the cellular environment. I'll review different approaches to inferring such 'regulatory codes' and then I'll focus on a statistical method, where codes are inferred using gene expression data for thousand of genes, dozens of different conditions and multiple species...more >
Monday, 06.11.12, 11 AM - 12 PM, Jacobs Hall, Room 4309
Data Compression and Secrecy
Prakash Narayan, University of Maryland, College Park
The multiterminal data compression problem of attaining omniscience and the secrecy problems of secret key generation and secure computing might suggest contrasting communication requirements. In fact, they are innately coupled. In this talk, we discuss connections between omniscience attainment by multiple terminals which observe separate but correlated signals, and secret key generation and secure function computation by those terminals, all in a distributed manner...more >
Wednesday, 05.30.12, 11 AM - 12 PM, Jacobs Hall, Room 4309
On Equivalence, Dependence, and Delay: Results from a Simple Tool for Information Theory
Michelle Effros, Caltech
The expansion of information theory from the study of very small networks to the understanding of extremely large networks is often viewed as both critically important and insurmountably difficult. Nonetheless, many general properties of large networks can be derived using very simple tools. This talk focuses on a reduction strategy borrowed from CS theory, exploring a few simple applications and their implications for understanding the nature of noise, the impact of dependence, and the consequences of delay for reliable communications in large (and small) communication networks...more >
Friday, 04.20.12, 11 AM - 12 PM, Jacobs Hall, Room 2512 (Booker Conference Room)
Optimal Power Flow and Demand Response
Steven Low, Caltech
Optimal power flow (OPF) problems determine the most efficient power generations, reactive powers for voltage support, or demand response. They are well-known to be nonconvex and hence NP hard. In the first part of the talk, we propose exact relaxations that can be solved efficiently. We prove that, if the network is radial (tree), then a semidefinite relaxation of OPF is always exact, provided the constraints on power flows satisfy a simple pattern. Using the branch flow model, we propose another simple conic relaxation to OPF, and prove that it is exact for radial networks. For mesh networks, we provide an exact condition under which the relaxation can is exact. We propose a simple method to convexify a mesh network using phase shifters so that OPF for the convexified network can always be solved efficiently for a globally optimal solution. We prove that convexification requires phase shifters only outside a spanning tree of the network graph.

In the second part of the talk, we describe a simple model that integrates two-period electricity markets, uncertainty in renewable generation, and real-time dynamic demand response. A load serving entity decides its day-ahead procurement to optimize expected social welfare a day before energy delivery. At delivery time when renewable generation is realized, it coordinates with users, in a decentralized manner, to manage load and purchase real-time balancing power in the real-time market, if necessary. We derive the optimal day-ahead decision, propose real-time demand response algorithm, and study the effect of volume and variability of renewable generation on the optimal social welfare.

(Joint work with Subhomesh Bose, Mani Chandy, Masoud Farivar, Dennice Gayme, Libin Jiang, Javad Lavaei, Caltech, and Chris Clarke, SCE.)

Friday, 04.15.11, 1:00 p.m. - 2:00 p.m., Jacobs Hall, Booker Conference Room, 2512
Network architecture and its discontents
John Doyle, Caltech
Measurement, prediction, communication, computation, decision, and control are at the heart of the modern mathematical theories in engineering and science. Yet research in these areas remains largely fragmented, and at times incompatible. The principle aim of this talk is to motivate an integrated theory based on optimization and robustness. Insights will be drawn from three converging research themes...more >
Tuesday, 01.18.11, 11:00 a.m. - 12:00 p.m., Jacobs Hall, Booker Conference Room, 2512
Sampling in the Age of Sparsity
Martin Vetterli, Ecole Polytechnique Fédérale de Lausanne, Switzerland and UC Berkeley
Joint work with Thierry Blu (CUHK), Y.Barbotin, Ali Hormati (EPFL), Y.Lu (Harvard), Pier-Luigi Dragotti (ICL) and Pina Marziliano (NTU).

Sampling is a central topic not just in signal processing and communications, but in all fields where the world is analog, but computation is digital. This includes sensing, simulating, and rendering the real world...more >
Friday, 10.29.10, 1:30 p.m. - 2:30 p.m. , Jacobs Hall, Booker Conference Room, 2512
Machine Learning Challenges in Ecological Science and Ecosystem Management
Thomas G. Dietterich, Oregon State University
Just as machine learning has played a huge role in genomics, there are many problems in ecological science and ecosystem management that could be transformed by machine learning. This talk will give an overview of several research projects at Oregon State University in this area and discuss the novel machine learning problems that arise. These include (a) automated data cleaning and anomaly detection in sensor data streams, (b) automated classification of images of arthropod specimens, (c) species distribution modeling including modeling of bird migration from citizen science data, and (d) design of optimal policies for managing wildfires in forest ecosystems. The machine learning challenges include flexible anomaly detection for multiple data streams, trainable high-precision object recognition systems, explicit models of sampling bias and measurement processes, and optimization of complex spatio-temporal Markov processes.
Friday, 05.07.10, 11:00 a.m. - 12:00 p.m., EBU-l, Room 2512
Real-Time Embedded Convex Optimization
Stephen P. Boyd, Stanford University
This talk concerns the use of convex optimization, embedded as part of a larger system that executes automatically with newly arriving data or changing conditions, in areas such as automatic control, signal processing, real-time estimation, real-time resource allocation and decision making, and fast automated trading. Such systems are already in use in applications such as model predictive control or supply chain optimization, with sample times measured in minutes (or longer); our focus is on systems with much faster dynamics, with execution times measured in milliseconds or microseconds for small and medium size problems. We describe a preliminary implementation of an automatic code generation system, which scans a description of the problem family and performs much of the analysis and optimization of the algorithm, such as choosing variable orderings used with sparse factorizations, at code generation time; compiling the generated source code yields an extremely efficient custom solver for the problem family.
Friday, 05.22.09, 3:00 PM  - 4:00 PM, EBU-l, Room 2512
On Optimal Fix-free Codes
Serap Savari, Texas A&M University
Fix-free codes are variable length codes in which no codeword is the prefix or suffix of another codeword. They have been investigated for joint source-channel coding and have been applied within the video standards H.263+ and MPEG-4 because their property of efficient decoding in both the forward and backward directions assists with error resilience. They are also interesting for problems in information retrieval such as searching for patterns directly in compressed text. We provide a low-complexity heuristic to produce fix-free codes. The design of optimal or minimum-redundancy fix-free codes has been a longstanding open problem. We offer the first solution both to this problem and to a variation in which all codewords are also required to be palindromes.
Friday, 05.15.09, 11:00 a.m. - 12:30 p.m. , EBU-l, Room 2512
Can MEMS Revolutionize the Semiconductor Industry? Sensor Integration, Energy Scavenging, and Tunable Radio
Farrokh Ayazi, Georgia Institute of Technology
Recent years have witnessed the maturity of MEMS in automotive safety applications and its penetration into cost sensitive consumer markets. Breakthrough researches have enabled high performance silicon-based sensors, actuators and signal-processing elements such as accelerometers, gyroscopes, microphones, micromirrors, high-Q microresonators and switches...more >
Friday, 05.08.09, 1:00 p.m. - 2:00 p.m., EBU-l, Room 2512
How Random-Set Theory Can Help Wireless Communications
Dr. Ezio Biglieri, UCLA
A number of problems arising in the analysis of wireless communication systems require the estimation of the components of a vector whose dimension is unknown. Two of these problems are, (a) Multiuser detection in which the number of active users is unknown and time-varying, and (b) Estimation of multipath channels where the number of paths is not known a priori and possibly time-varying. Standard solutions to these problems are intrinsically suboptimum, as they proceed either by assuming a fixed number of vector components, or by first estimating this number, then the values taken on by the components. For an optimum solution, one may use random-set theory, which is a probability theory of finite sets exhibiting randomness not only in their elements, but also in their cardinality. Some examples, related to problems (a) and (b) above, illustrate the results obtained from this theory.
Friday, 05.01.09, 1:30 p.m. - 2:30 p.m. , EBU-l, Room 2512
On the intersection between queueing and information theory - some new results.
Muriel Medard, MIT
The interaction between queueing and information theory is an important one in bringing information-theoretic principles into networking. In this talk, we highlight some new results in this area, with an emphasis on wireless systems. The first set of results consider the issue of coding in queues, in particular in the context of network coding. Coding in this context blurs the traditional delineation among packets, but it allows for considerable benefits for delay...more >
Friday, 02.27.09, 3:00 PM - 4:15 PM, Booker Conference Room, EBU 1 Room 2512
Patrick Yue, UC Santa Barbara
Friday, 02.15.08, 11, TBD
The power and weakness of computational randomness
Avi Wigderson, Institute for Advanced Study
Man has grappled with the meaning and utility of randomness for centuries.
Research in the Theory of Computation in the last thirty years has enriched
this study considerably. I will describe two main aspects of this research on
randomness, demonstrating its power and weakness respectively.

-Randomness is paramount to computational efficiency:

The use of randomness can dramatically enhance computation (and do other
wonders) for a variety of problems and settings...more >
Friday, 11.30.07, 11 - 12, EBU-1 4307
Wireless Networks: Cellular System Engineering and Multiterminal Information Theory
Pramod Viswanath, UIUC
Successful design of reliable networks, for information compression and transmission, requires posing questions at appropriate levels of abstraction. At a coarser level, one asks for optimal architectures. At a finer level, once a specific architecture has been adopted, one asks for efficient engineering schemes. This talk discusses two pieces of work that exemplify this approach, one in the context of information transmission and the other for information compression...more >
Friday, 11.02.07, 11 - 12, EBU-1 4307
Extracting Information from Music Audio
Dan Ellis , Columbia University
Large online collections of music audio present a rich opportunity for automatic analysis, particularly by machine learning techniques that can take advantage of huge datasets. We have been investigating a number of music audio analysis applications, mostly oriented towards the goal of prediction of perceived similarity or preference. I will describe several of these projects, including music classification based on spectral feature distributions, incremental learning of playlist classes, rhythm pattern decomposition, and automatic detection of cover song versions...more >
Friday, 10.05.07, 11 - 12, EBU-1 4307
Wojciech Szpankowski, Purdue University
Information permeates every corner of our lives and shapes our universe. Understanding and harnessing information holds the potential for significant advances. The breadth and depth of underlying concepts of the science of information transcend traditional disciplinary boundaries of scientific and commercial endeavors. Information can be manifested in various forms: business information is measured in dollars; chemical information is contained in shapes of molecules; biological information stored and processed in our cells prolongs life...more >
Friday, 05.05.06, 11:00 am - 12:00 pm, Calit2 Multipurpose Room
Capacity, cooperation, and cross-layer design in wireless wetworks
Andrea Goldsmith, Stanford University
We consider fundamental capacity limits in wireless networks where nodes can cooperate in transmission, reception, and routing. We propose novel cooperation techniques that approach the capacity bounds, including virtual MIMO transmission, relaying, and conferencing. The optimal cooperation strategy is cross-layer in nature and depends on the network topology, the channel SNR, and the channel side information available at the nodes...more >
Friday, 04.21.06, 4:00 p.m., CMRR Auditorium
Shannon lecture: Demodulation Meets Signal Processing - Two-Dimensional Information Theory
Richard E. Blahut, University of Illinois at Urbana-Champaign
This year's Shannon Memorial Lecture by Professor Blahut will examine agorithms for demodulation and signal processing such as the Viterbi algorithm and the Wiener filter are highly valued by those who use them. Motivated by problems of two-dimensional recording, we will consider these problems in a unified way. The Richardson-Lucy algorithm will be described as an information-theoretic algorithm suitable for both demodulation and signal processing, and often superior to both the Viterbi algorithm and the Wiener filter (work based in part on the thesis of Zhijun Zhao...more >
Friday, 03.17.06, 11:00 - 12:00, Calit2 MPR
Signal Processing for Genomics
P.P. Vaidyanathan, Caltech
The theory and methods of signal processing are becoming increasingly important in molecular biology. This talk includes a brief review of molecular biology, followed by a review of the applications of signal processing theory. Also presented is the relatively new topic of noncoding genes, and the associated problem of identifying ncRNA buried in DNA sequences...more >
Friday, 03.03.06, 11:00 - 12:00, Calit2 MPR
Error-Correcting Codes for Automatic Control
Leonard Schulman, Caltech
Increasingly, systems with automatic feedback control may consist of several remote devices, connected only by unreliable communication channels. It is necessary in these conditions to have a method for accurate, real-time state estimation in the presence of channel noise.

We address this problem, for the case of polynomial-growth-rate state spaces, by providing a new type of error-correcting code that is on-line and computationally efficient...more >
Friday, 02.17.06, 11:00 - 12:00, Calit2 MPR
Design in the Late Silicon Age
Jan Rabaey, UC Berkeley
Scaling of silicon integrated technology into the deep sub-100 nm space brings with it a number of formidable challenges to the designer. Issues such as design complexity, power dissipation, process variability and reliability are challenging the traditional design methodologies. In this presentation, it is conjectured that the only viable long-term solution to these challenges is to drastically revise the way we do design, and a roadmap of potential solutions is presented...more >
Friday, 02.03.06, 11:00 - 12:00, Calit2 MPR
Approaches to Network Coding
Ralf Koetter, UIUC
Network coding allows information from different streams to be processed or coded together at network nodes. This is a more general and powerful framework than the traditional store and forward approach. This talk will give a short introduction to this field and its ramifications. Particular attention will be paid to the problem of actually designing network codes in the multiple unicasts scenario in directed graphs...more >
Friday, 01.27.06, 11:00 - 12:00, Calit2 MPR
Quantization, Compression, and Classification: Extracting Discrete Information from a Continuous World
Robert M. Gray
Scientists and engineers often seek to measure, communicate, store, process, reproduce, or analyze signals encountered in the real world. Most such signals are inherently continuous or analog in nature, yet increasingly the means for communicating, storing, and processing such information are discrete. Generally something is lost when continuous information is converted into discrete approximations, so a natural goal is to preserve as much of the original information as possible...more >
Friday, 01.20.06, 11:00 - 12:00, Calit2 MPR
MM-wave RFIC
Hossein Hashemi
In this talk, I will present an overview of various forms of multi-antenna configurations such as phased-arrays, beam forming schemes, and diversity multiple-input multiple-output (MIMO) systems. Antenna arrays for communications, radar, imaging, and sensing applications will be described in more detail. I will then describe our current research in the following three areas:

Ultra-Wide Band (UWB) beam-forming on silicon for high resolution ranging applications

Silicon-based integrated phased arrays at millimeter waves for high data rate communication and imaging applications

Concurrent multi-frequency beam-forming for multi-functional communication/imaging/sensory systems

Various transceiver architectures and circuit implementations for a complete integration of multi-antenna systems using a silicon technology and experimental results will be presented...more >