Once every week while school is in session, EECS invites a distinguished researcher or practitioner in a computer science or electrical and computer engineering-related field to present their ideas and/or work. Talks are generally targeted to electrical engineering and computer science graduate students. This colloquium series is free and open to everyone.

Upcoming Colloquia

Efficient Inference Methods for Probabilistic Logical Models

KEC 1001

Speaker Information

Sriraam Natarajan
Department of Computer Science
University of Wisconsin, Madison

Abstract

Probabilistic Logical Models (PLMs) combine the powerful formalisms of probability theory and first-order logic to handle uncertainty in large, complex problems. While PLMs provide a very effective learning paradigm under the umbrella of Statistical Relational Learning (SRL) methods, tractable inference is a significant problem in these models. Earlier approaches focused on grounding the model to a propositional network to use existing inference algorithms. Other popular techniques include sampling and lifted inference, with a lot of interest in the latter recently.

In this talk, I will present three different approaches to accelerate inference in PLMs. First, a preprocessing method for Markov Logic Networks that makes exact grounded inference tractable; second, an approximate inference method called `counting belief propagation' that performs belief propagation on compressed factor graphs; and finally, an `anytime'inference algorithm that returns a bound over the marginal distribution of the query variable. I will present experimental results to demonstrate the usefulness of these three distinct, yet related, inference methodologies.

Speaker Bio

Sriraam Natarajan is currently a Post-Doctoral Research Associate at the Department of Computer Science at University of Wisconsin-Madison. He graduated with his PhD from Oregon State University working with Dr. Prasad Tadepalli. His research interests lie in the field of Artificial Intelligence, with emphasis on Machine Learning, Statistical Relational Learning, Reinforcement Learning, Graphical Models and Bio-Medical Applications.

Provably good codes for secure hash function design

KEC 1001

Speaker Information

Anindya Patthak
Computational Lithography Group
Intel

Abstract

n this talk, we describe a technique to give a lower bound on the minimum distance of certain types of quasi-cyclic codes. The key idea is to reduce this problem to prove a lower bound on the minimum distance of a few significantly smaller codes. Using this technique, we show that a code which is similar to SHA-1 (Secure Hash Algorithm) message expansion code, has a minimum distance of 82. The technique is further applied to prove that the minimum weight of the SHA-1 message expansion code is small. The low minimum distance of SHA-1 message expansion code has earlier been exploited to cause an effective collision attack on SHA-1. Estimating the minimum distance of a code given by its parity check matrix is well known to be a hard problem. This technique is expected to be helpful in estimating the minimum distance of similar quasi-cyclic codes as well as in designing future practical cryptographic hash functions.

Speaker Bio

Anindya C. Patthak presently works for Intel Corporation in the Computational Lithography Group. Earlier he was a post doctoral researcher at the University of California, Riverside. He received his Ph.D. from the University of Texas at Austin in 2007, under the guidance of Prof. David Zuckerman, and a B.Tech. from IIT Kharagpur. His research focuses on error correcting codes and their applications to theoretical computer science. He is also interested in algebraic property testing, algebraic-combinatorial constructions and pseudo-randomness etc.

Talkative Sensors: Collaborative Machine Learning for Volcano Sensor Networks

KEC 1001

Speaker Information

Kiri Wagstaff
Senior Researcher
NASA Jet Propulsion Lab

Abstract

Imagine a machine learning agent deployed at each station in a sensor network, so that it can analyze incoming data and determine when something interesting happens. Traditionally, this analysis would be done independently at each station.  But what if each agent could talk to its neighbors and find out what they're seeing? We’ve developed a learning system that enables collaboration so that the agents can autonomously (without human input) improve their performance. Each agent can ask its neighbors for their opinions, then use them to refine its own results. When each agent is given the task of clustering the observed data, the opinions are expressed in the form of pairwise clustering constraints. We evaluated several heuristics for selecting which items an agent should query and found that the best strategy was to select one item close to its assigned cluster and one item at the boundary between two clusters. We applied this technique to seismic and infrasonic data collected by the Mount Erebus Volcano Observatory, in which the goal was to separate eruptions from non-eruptions. Collaborative clustering achieved a 150% improvement over regular, non-collaborative clustering. This is joint work with Jillian Green (California State Univ., Los Angeles), Rich Aster and Hunter York (New Mexico Institute of Mining and Technology), Terran Lane (Univ. of NM), and Umaa Rebbapragada (JPL), funded by the NSF.

Speaker Bio

Kiri Wagstaff is a senior researcher in artificial intelligence and machine learning at the Jet Propulsion Laboratory, on sabbatical at Oregon State University for the fall term. Her focus is on developing new machine learning and data analysis methods, particularly those that can be used for in situ analysis onboard spacecraft (orbiters, landers, etc.). She has developed several classifiers and detectors for data collected by instruments on the EO-1 Earth orbiter, Mars Pathfinder, and Mars Odyssey. The applications range from detecting dust storms on Mars to predicting crop yield on Earth. She holds a Ph.D. in Computer Science from Cornell University (2002) and an M.S. in Geological Sciences from the University of Southern California (2008).

High-Frequency Limits of Wireless and Wireline Circuits in Silicon Processes

KEC 1001

Speaker Information

Dr. James Buckwalter
Assistant Professor
Electrical and Computer Engineering
University of California, San Diego

Abstract

This talk will describe how advanced CMOS processes are changing the landscape of wireless and optical communication technologies. Emerging millimeter-wave applications require antenna arrays with RF front-ends that need additional functionality without sacrificing power efficiency. Our group has recently reported several novel circuit techniques including constructive wave amplifiers to realize bidirectional millimeter-wave front-ends and efficient power amplifiers in 45-nm Silicon-on-Insulator (SOI) CMOS.

Additionally, high-speed (>10 Gb/s) silicon photonic circuitry for chip-to-chip communication will be discussed.  New optical wireline systems are proposed through the development of photonic devices compatible with standard silicon processes. Our research group is investigating channel coding for silicon microring resonators in 130-nm SOI CMOS and approaches to reduce power consumption of 40-Gb/s circuitry in 45-nm SOI CMOS.

Speaker Bio

Prof. James Buckwalter supervises the high-speed integrated circuits laboratory at the University of California - San Diego. His research interests are RF and millimeter-wave chip design for wireless applications and opto-electronic interface circuitry. His research has been recognized with the DARPA Young Faculty Award in 2007 and NSF Career Award in 2011.

Smart Home Technologies

KEC 1001

Speaker Information

Dr. Aaron S. Crandall
Postdoctoral Research Associate
Washington State University
Center for Advanced Studies in Adaptive Systems

Abstract

The population of the United States is aging. By 2040 the largest age group will be 80-plus. We do not have the care facilities or care providers to handle this upcoming wave of older adults. New approaches and tools are needed to address this issue. There is a movement among the gerontology field to move towards an aging in place philosophy, where people need to live in their homes longer instead of moving to a care facility. The CASAS group at WSU has been focused on building smart home technologies that can assist residents and care givers with this aging in place philosophy.

This talk on smart home technologies will introduce the CASAS research environment and the wide range of applications that these smart homes are used for. The CASAS group has constructed a series of smart home testbeds in private homes for real world data collection. This has enabled them to explore algorithms for interpreting and mining smart home data using a wide range of artificial intelligence, machine learning and data mining strategies. The goal is to enable computers to build accurate models of what and how well the smart home residents are doing in their day to day activities. With these new capabilities, systems to help residents live in their homes longer become more feasible.

Speaker Bio

Dr. Crandall is a postdoctoral research associate at Washington State University's Center for Advanced Studies in Adaptive Systems (CASAS). His work has included the application of behaviometrics to smart home systems, as well as techniques for handling multiple residents within smart homes. These tools are geared towards the overall goals of the CASAS smart home research group to build novel approaches to support aging in place and eldercare technologies. Dr. Crandall received his Ph.D. in Computer Science from Washington State University in 2011, his Master of Computer Science from Oregon Health and Science University in 2006 and his Bachelor of Electrical Engineering from the University of Portland in 2001

Analog Impairments Correction of Quadrature Mixers in Stand Alone Receivers

KEC 1001

Speaker Information

Radu Fetche
IC Design Engineer
Rohde and Schwarz USA

Abstract

In a complex radio communication system most of the radio’s impairments are digitally corrected, assisted by the digital baseband processor/controller. There are applications where the radio is placed on a separate die from the digital baseband processor for performance or business related reasons. These stand alone radios are analog in nature with digital content only in the ADC, baseband filtering and PLL.

This talk will focus on the techniques used for correction of the quadrature mixer impairments in the analog domain. Several options are presented to correct for phase and amplitude imbalance, second and third order nonlinearity and the local oscillator leakage.

Speaker Bio

Radu Fetche is an IC design engineer at Rohde and Schwarz USA in Beaverton OR. Prior to joining Rhode and Schwarz USA, he was a senior design engineer with the TV and Radio Group of MAXIM Integrated Products in Hillsboro OR. His research focuses on high-performance microwave and radio circuits for communications and measurement applications.

Virtual Separation of Concerns: Toward Preprocessors 2.0

KEC 1001

Speaker Information

Christian Kästner
Post-Doctoral Researcher
Group for Programming Languages and Software Engineering
Philipps University Marburg, Germany

Abstract

Despite much research on sophisticated programming languages and generators to implement software product lines, practitioners still implement variability in most industrial product lines with the   preprocessor. By annotating code fragments with #ifdef and #endif directives, different programs with or without these fragments can be created. However, despite their common use in practice, preprocessors are heavily criticized for their negative effect on code quality and maintainability: Preprocessors neglect separation of concerns, are prone to introduce subtle errors, and can entirely obfuscate the source code. We discuss how to tame preprocessor variability. We show how already simple tool support, including views and visualizations, emulates separation of concerns virtually, eliminates common pitfalls, and supports developers in program understanding tasks. Additionally, we explore strategies to detect and prevent errors that show up only in specific configurations. To that end, we have developed a variability-aware type system that can check entire product lines without generating all configurations. We report experience from scaling our tools to the Linux kernel with 6 million lines of code and 10000 configurations options. Our goal is to parse and type checking all up to 2^10000 configurations of the Linux kernel in one pass.

Speaker Bio

Christian Kästner is a PostDoc at the Group for Programming Languages and Software Engineering of the Philipps University Marburg, Germany. He received his Ph.D in Computer Science in 2010 from the University of Magdeburg, Germany for his work on virtual separation of concerns. For his dissertation, he received the best-dissertation award from the German Computer Science Association (GI). His research focuses on correctness and understanding of systems with variability, including work on implementation mechanisms, tools, variability-aware analysis, type systems, feature interactions, and refactoring. He is the author or coauthor of over a fifty peer-reviewed scientific publications.

Challenges to Energy Efficient Mobile Platforms

Cordley 2113

Speaker Information

Dr. Tawfik Arabi
Senior Principal Engineer
Mobility Group
Intel Corp

Abstract

Delivering power efficiently to advanced technology VLSI is a top challenge and priority for mobile platforms. In some cases, near 50% of the real estate on a mobile platform area is used to convert power to the low voltages required by the advanced technology. This is so because as voltage continues to scale down with technology, the tolerances required by these circuits for reliable and energy efficient operation continues to tighten. In this presentation we review the state of the art and show a large gap between today and the desired energy efficiency of computing. We will also show how integration of the power delivery circuitry presents an opportunity for non-linear improvements in energy efficiency of mobile devices. We will present Intel’s view of the required advancements in circuit technology, VLSI testing technology, and process technology needed in the next 5 to 10 years to achieve such integration.

Speaker Bio

Dr. Arabi is a senior principal engineer in the Mobility Group at Intel Corporation. He is responsible for defining low power technologies for the next generation notebook platform.

Dr. Arabi is the recipient of many Intel and IEEE awards including 6 Intel Achievement award (Intel’s most prestigious award given by the executive staff to only a handful of teams every year). Dr. Arabi is an IEEE Fellow and has over 100 Journal and conference publications. He supported and directed research programs between Intel and several leading universities in the US and abroad. He is the founder and director of Intel’s Middle East Energy Efficiency Research Initiative. Dr. Arabi Graduated from AUB in 1985 with a B.S.E.E and a Ph.D. from Syracuse University in 1991.

CMOS Switched-Capacitor Circuits: Recent Advances in Bio-Medical and RF Applications

KEC 1001

Speaker Information

David J. Allstot
Professor
Department of Electrical Engineering
University of Washington

Abstract

The switched-capacitor technique has been used in high-volume data conversion and signal processing ICs for more than three decades. It is also ubiquitous in RF transceiver circuits because it uses capacitors, which are area-efficient native devices in CMOS technologies, and MOSFETs operating as switches.
 
The RF power amplifier dissipates a large fraction of the total power of a transceiver because of its low efficiency. Despite more than two decades of intensive research, the challenge of on-chip RF PAs with high efficiency in digital-friendly CMOS technologies has not been met. Switching PA topologies with relatively high efficiency have gained momentum, and relatively high output power is being delivered using power combining techniques. Supply regulation techniques have enabled higher efficiency when amplifying non-constant envelope modulated signals. A new paradigm—the switched-capacitor RF power amplifier—which meets many of the remaining challenges is described.
 
Body-area-networks (BAN) that integrate multiple sensor nodes in portable and wearable bio-medical systems are revolutionizing healthcare. A typical BAN comprises several bio-signal and motion sensors and uses ultra-low-power short-haul radios in conjunction with nearby smart-phones or handheld devices (with GPS capabilities) to communicate via the internet with a doctor or other healthcare professional. Higher energy efficiency is critical to the development of feature-rich, wearable and reliable personal health-monitoring systems.
 
The amount of data transmitted to the smart-phone increases as more sensors are added to the BAN. Because the energy consumed for RF transmission is proportional to the data rate, it is advantageous to compress the bio-signal at the sensor prior to digitization and transmission. This energy-efficient paradigm is possible using compressed sensing—a new sampling theory wherein a compressible signal can be acquired using only a few incoherent measurements. For ECG signals, for example, compression factors up to 16X are achievable which means similar reductions in energy consumption. The second part of this talk will overview compressed sensing techniques and describe a switched-capacitor analog front-end for bio-signal acquisition.

Speaker Bio

David J. Allstot received the B.S. from the Univ. of Portland, the M.S. from Oregon State Univ. and the Ph.D. from the Univ. of California, Berkeley. He has held several industrial and academic positions and has been the Boeing-Egtvedt Chair Professor of Engineering at the Univ. of Washington since 1999. He was Chair of the Dept. of Electrical Engineering from 2004 to 2007. Dr. Allstot has advised approximately 100 M.S. and Ph.D. graduates, published about 300 papers, and received several awards. He has also been active in service to IEEE.

Prefab: What if anybody could modify any interface?

KEC 1001

Speaker Information

James Fogarty
Associate Professor
Computer Science and Engineering
University of Washington

Abstract

Widely-used interface toolkits currently stifle the progress and impact of user interface research. Advances are limited by both the rigidity of current interfaces and the fragmentation of applications among many different underlying toolkits. Many promising innovations therefore remain difficult or impossible to deploy.

Prefab examines pixels as a universal representation of the desktop. By reverse engineering the pixel-level appearance of interface elements, Prefab enables runtime interface modification without access to source and without cooperation from the underlying application. I will motivate this approach, present Prefab's pixel-based methods, and show several examples of runtime interface enhancements. This will include our implementation of the first general-purpose target-aware pointing enhancement, an idea proposed more than 15 years ago that has previously been considered impractical to actually deploy. I will conclude by discussing the potential of this work for catalyzing future interaction research and beginning to democratize our everyday interfaces.

Speaker Bio

James Fogarty is an Associate Professor of Computer Science & Engineering at the University of Washington. He is also a core member of the DUB Group, the University of Washington's cross-campus initiative advancing Human-Computer Interaction and Design research and education. His broad research interests are in Human-Computer Interaction, User Interface Software and Technology, and Ubiquitous Computing. His focus is on new methods and tools for building interactive software, including the challenges of developing, deploying, and evaluating intelligent technologies. He is co-chairing the CHI 2012 subcommittee for "Expanding Interaction through Technology, Systems & Tools", and he has received Best Paper awards or nominations at CHI 2010, UbiComp 2009, CHI 2009, and CHI 2005. His research has been generously supported by the National Science Foundation, FXPAL, Google, Intel, and Microsoft.