Wednesday, October 27, 2010

SESSION I
10:05 - 11:10
Security and Privacy
CSE 305
Ubiquitous Computing
CSE 403
Computation Aiding Biology and Biology Aiding Computation
CSE 691
SESSION II
11:20 - 12:25
Technology for Under-Served Regions
CSE 305
RGB-D: Advanced Reasoning with Depth Cameras
CSE 403
Solving Hard Real World Problems with Games
CSE 691
Keynote Talk
1:40-2:40
The UW Center for Commercialization: Tomorrow's University Now
Atrium


SESSION III
2:40 - 3:45
Software Development
CSE 305
Computers that Read
CSE 403
Battery-free and Low-Power Wireless Sensing
CSE 691
SESSION IV
3:55 - 5:00
HCI
CSE 305
Configurable Computing
CSE 403
Sustainability Sensing
CSE 691

Session I

    Security and Privacy (CSE 305)

    • 10:05-10:10: Introduction and Overview: Yoshi Kohno

    • 10:10-10:30: Experimental Security Analysis of a Modern Automobile, Alexei Czeskis

      Modern automobiles are no longer mere mechanical devices; they are pervasively monitored and controlled by dozens of digital computers coordinated via internal vehicular networks. While this transformation has driven major advancements in efficiency and safety, it has also introduced a range of new potential risks. We experimentally evaluate these issues on a modern automobile and demonstrate the fragility of the underlying system structure. We demonstrate that an attacker who is able to infiltrate virtually any Electronic Control Unit (ECU) can leverage this ability to completely circumvent a broad array of safety-critical systems. Over a range of experiments, both in the lab and in road tests, we demonstrate the ability to adversarially control a wide range of automotive functions and completely ignore driver input — including disabling the brakes, selectively braking individual wheels on demand, stopping the engine, and so on. We find that it is possible to bypass rudimentary network security protections within the car, such as maliciously bridging between our car's two internal subnets. We also present composite attacks that leverage individual weaknesses, including an attack that embeds malicious code in a car's telematics unit and that will completely erase any evidence of its presence after a crash. Looking forward, we discuss the complex challenges in addressing these vulnerabilities while considering the existing automotive ecosystem.

    • 10:30-10:50: Human Values and Directions for Securing Implantable Medical Devices, Tamara Denning(PDF Slides)

      Implantable medical devices (IMDs) improve patients' quality of life and help sustain their lives. In this study, we explore patient views and values regarding their devices to inform the design of computer security for wireless IMDs. We interviewed 13 individuals with implanted cardiac devices. Key questions concerned the evaluation of 8 mockups of IMD security systems. Our results suggest that some systems that are technically viable are nonetheless undesirable to patients. Patients called out a number of values that affected their attitudes towards the systems, including perceived security, safety, freedom from unwanted cultural and historical associations, and self-image. In our analysis, we extend the Value Sensitive Design value dams and flows technique in order to suggest multiple, complementary systems; in our discussion, we highlight some of the usability, regulatory, and economic complexities that arise from offering multiple options. We conclude by offering design guidelines for future security systems for IMDs.

    • 10:50-11:10: Seeing Through Obscure Glass, Qi Shan(PDF Slides)

      Obscure glass is textured glass designed to separate spaces and 'obscure' visibility between the spaces. Such glass is used to provide privacy while still allowing light to flow into a space, and is often found in homes and offices. We propose and explore the challenge of 'seeing through' obscure glass, using both optical and digital techniques. In some cases — such as when the textured surface is on the side of the observer — we find that simple household substances and cameras with small apertures enable a surprising level of visibility through the obscure glass. In other cases, where optical techniques are not usable, we find that we can model the action of obscure glass as convolution of spatially varying kernels and reconstruct an image of the scene on the opposite side of the obscure glass with surprising detail.

    Ubiquitous Computing (CSE 403)

    • 10:05-10:10: Introduction and Overview: Shwetak Patel

    • 10:10-10:30: HeatWave: Thermal Imaging for User Interaction on Arbitrary Surfaces, Eric Larson (PDF Slides)

      Thermal sensing has had limited examination in the HCI research community and is generally under-explored outside of law enforcement and energy auditing applications. Our work examines the role of thermal imaging as a new sensing solution for enhancing user surface interaction. In particular, we demonstrate how thermal imaging and existing computer vision techniques can make segmenting and detecting routine interaction techniques possible in real-time and complement or simplify algorithms for traditional RGB and depth cameras. Example interactions include (1) distinguishing surface touch or target selection from hovering above surface, (2) shape-based gestures similar to ink strokes, (3) pressure based gesturing, and (4) multi-finger surface-based gestures. We close by discussing the practicality of thermal sensing for naturalistic user interaction and future work.

    • 10:30-10:50: Mobile Haptic Feedback, Sidhant Gupta and Francis Iannacci(PDF Slides)

      Three novel mobile haptic feedback devices will be presented. The first device, SqueezeBlock, is a haptic device that conveys information by changing its stiffness. For example, it can dynamically become easy or stiff to squeeze depending on the battery level of your phone. Our second device, called InGen, is a self-powered haptic feedback system that uses a single motor to generate power, perform sensing, and generate haptic feedback. The final device which we will introduce, Haptic Laser, is a multi-sensation tactile feedback system for exploring and interacting with physical objects at a distance.

    • 10:50-11:10: Mobile Lung Function: Counting Coughs and Spirometry on a Mobile Phone, Eric Larson(PDF Slides)

      Health aspects on the mobile phone have become particularly pervasive topics - especially when using only the existing microphone as the medical sensor. As such, we present a mobile phone-based sensing algorithm for detecting and counting a person's coughs and cough epochs (or episodes) throughout the day in a naturalistic environment. Existing approaches to cough assessment either require a patient to self-monitor their coughs or require wearing specialized equipment. We recruited 17 participants experiencing cough episodes, and gave each of them a commodity mobile phone to record a continuous audio stream of themselves and their environment. In all, we recorded 72 hours of audio and hand-annotated 2,542 individual cough events for ground truth (1,016 cough epochs). We use this corpus to inform and evaluate algorithms for detecting and classifying cough sounds with an average accuracy of 88.4% using different subjects for training and testing. Additionally, we discuss how a mobile phone microphone might be leveraged to act as an additional medical instrument - such as an ultra low cost spirometer.

    Computation Aiding Biology and Biology Aiding Computation (CSE 691)

    • 10:05-10:10: Introduction and Overview: Martin Tompa

    • 10:10-10:30: Computation Helps Tell Us How Muscles Are Made, Walter L. Ruzzo(PDF Slides)

      In the textbooks, genes are controlled by transcription factors, proteins that bind to DNA near the starts of genes to activate or repress them. One of the best-studied transcription factors, MyoD, is a master regulator of myogenesis, the process by which precursor cells differentiate into skeletal muscle cells. Indeed, expressing this one protein in a skin cell causes extensive remodeling of the cell, resulting in a facsimile of a muscle cell. Working with collaborators at the Fred Hutchinson Cancer Research Center, we searched for MyoD binding sites in differentiating and mature muscle cells. As expected, MyoD is found near start sites of the several hundred genes that are known to be driven by this powerful regulator. Surprisingly, however, these sites comprise only a tiny fraction of the sites bound by MyoD, many thousands of which occur far from any gene. These data suggest that MyoD is unexpectedly multifunctional, initiating changes in state that result in broad reprogramming of the cell.

    • 10:30-10:50: DNA Nanotechnology and Molecular Programming, Georg Seelig

      I will give a brief overview of the fields of DNA nanotechnology and molecular programming. In particular, I will focus on a few different ways in which computation can be embedded in molecular interactions, and will talk about the goals and potential applications for this research.

    • 10:50-11:20: Foldit, an Online Game for Protein Structure Prediction, Zoran Popović

      Predicting a stable three-dimensional structure from any given amino acid sequence using first physical principles remains a formidable computational challenge. Aiming to recruit human visual and strategic powers to the task, we created an online multiplayer game called Foldit, in which thousands of nonscientists compete and collaborate to produce a rich set of search strategies for protein structure refinement. This work, published in the August 2010 issue of Nature, shows that even computationally complex scientific problems can be effectively crowd-sourced using interactive multiplayer games.

Session II

    Technology for Under-Served Regions (CSE 305)

    • 11:20-11:25: Introduction and Overview: Gaetano Borriello

    • 11:25-11:40: FoneAstra: A Low-cost, Networked Sensing Platform for Developing Regions, Rohit Chaudhuri (PDF Slides)

      FoneAstra is a low-cost, programmable device that extends capabilities of mobile phones. We discuss how our device extends the functionality of non-programmable, low-tier mobile phones that are most prevalent amongst low-income groups in developing regions. FoneAstra enables interesting mobile applications in a variety of domains ranging from participatory sensing to remote monitoring to healthcare. We present an application in which FoneAstra is being used to monitor the temperature of vaccines in a country-wide "vaccine cold-chain." We also discuss an application, currently under development, in which FoneAstra will be used to monitor the milk pasteurization process at Human Milk Banks. Both of these applications are being done in collaboration with PATH, our Seattle-based partner NGO, that works in the healthcare delivery area for developing countries.

    • 11:40-11:55: Portable Ultrasound for Midwives, Waylon Brunette(PDF Slides)

      Ultrasound imaging is an effective tool for identifying maternal mortality risk factors. Unfortunately, ultrasound is nearly absent in many rural healthcare facilities in developing regions due to the high costs of both equipment and required training. To leverage existing healthcare systems commonly found in these contexts, we have focused our efforts on increasing the diagnostic capabilities of midwives –often central medical figures in rural and low-income communities. We have designed and built a low-cost portable ultrasound device consisting of a USB ultrasound probe and a touchscreen netbook that has a simplified user interface that specifically targets midwives. Our modular design approach allows for easy modification, and the device is designed to utilize existing local healthcare resources in order to create a sustainable solution that does not depend on continuous foreign assistance.

    • 11:55-12:10: Smart Connect: A Communication Link for Peripheral Health Facilities, Nell O'Rourke (PDF Slides)

      Rural health facilities in developing countries collect valuable data relating to patient care, however it is challenging and time consuming to report this data to the urban facilities where it could be aggregated and utilized. To address this problem, we are developing Smart Connect, a facility-based communication device that uses SMS messages to provide a data link between peripheral health facilities and a server connected to the Internet. In collaboration with PATH, we have completed fieldwork in Nicaragua that indicates that a variety of processes could be improved with the use of this device. These include filing epidemiological surveillance reports, receiving results of diagnostic test, and providing automatic monitoring of vaccine refrigeration equipment. In this talk, I will discuss the requirements gathering process we completed to learn about data collection in the Nicaragua health system, and I will describe areas in which Smart Connect can be used to improve data connectivity.

    • 12:10-12:25: Technology for Video-Based Agricultural Education in Rural India, Natalie Linnell(PDF Slides)

      This talk will discuss work exploring the use of handheld technology to assist mediators in facilitated video deployments. Facilitated video is a teaching method where a semi-skilled facilitator shows videos of excellent educators to students, pausing the video for questions and interaction. We partnered with Digital Green, an NGO which uses facilitated video for agricultural education in rural India. Based upon an investigation we conducted into the information needs of the facilitators, we built and field-tested two different solutions for delivering information to the facilitator in real time during the video shows. The primary difference between the two was the manner in which the device discovered which video was playing and its timing offset. One approach involved embedding audio codes into the video that were decoded on an Android smart phone using digital signal processing, while the other approach employed a custom-hardware "smart" remote control. We field tested both devices for four weeks in India. User response was very positive, and both approaches were shown to be viable for use in the field.

    RGB-D: Advanced Reasoning with Depth Cameras (CSE 403)

    • 11:20-11:25: Introduction and Overview: Xiaofeng Ren

    • 11:25-11:45: RGB-D mapping: Building Dense 3D Maps of Indoor Environments, Peter Henry (PDF Slides)

      RGB-D cameras are novel sensing systems that capture RGB images along with per-pixel depth information. We investigate how such cameras can be used in the context of robotics, specifically for building dense 3D maps of indoor environments. Such maps have applications in robot navigation, manipulation, semantic mapping, and telepresence. We present a full 3D mapping system that utilizes a novel joint optimization algorithm combining visual features and shape based alignment. We describe how we detect and optimize over loop closures.

      Our dense 3D maps can be efficiently represented as surfels, which are small surface patches allowing for efficient updates and occlusion reasoning. We demonstrate that it is possible to build extremely rich 3D maps with RGB-D cameras, suggesting that it is feasible to build complete robot navigation and interaction systems solely based on inexpensive depth cameras.

    • 11:45-12:05: Enabling Robots to Autonomously Build 3D Models of Objects, Mike Krainin(PDF Slides)

      Recognizing and manipulating objects is an important task for mobile robots performing useful services in everyday environments. While existing techniques for object recognition related to manipulation provide very good results even for noisy and incomplete data, they are typically trained using data generated in an offline process. As a result, they do not enable a robot to acquire new object models as it operates in an environment. We present an approach to building 3D models of unknown objects based on a depth camera observing the robot’s hand while moving an object. The approach integrates both shape and appearance information into an articulated ICP approach to track the robot's manipulator and the object. Additionally, we introduce an information-gain based variant of the next best view algorithm in order to determine how the manipulator should move the object in front of the camera. Experiments show that our approach provides very good 3D models even when the object is highly symmetric and lacking visual features and the manipulator motion is noisy.

    • 12:05-12:25: Gesture Recognition and Segmentation, Hao-Wei Liu (PDF Slides)

      With the availability of 3D cameras, depth information makes many vision problems easier to solve. In this talk, I will show that simple depth cues can be used for gesture recognition. I will also address the noise that come with the camera and possible ways to handle them.

    Solving Hard Real World Problems with Games (CSE 691)

    • 11:20-11:25: Introduction and Overview: Zoran Popović

    • 11:25-11:45: Reconstructing the World in 3D: Bringing Games with a Purpose Outdoors, Kathleen Tuite (PDF Slides)

      We are interested in reconstructing real world locations as detailed 3D models, but to achieve this goal, we require a large quantity of photographic data. We designed a game to employ the efforts and digital cameras of everyday people to not only collect this data, but to do so in a fun and effective way. The result is PhotoCity, a game played outdoors with a camera, in which players take photos to capture flags and take over virtual models of real buildings. The game falls into the genres of both games with a purpose (GWAPs) and alternate reality games (ARGs). Each type of game comes with its own inherent challenges, but as a hybrid of both, PhotoCity presented us with a unique combination of obstacles. This paper describes the design decisions made to address these obstacles, and seeks to answer the question: Can games be used to achieve massive data-acquisition tasks when played in the real world, away from standard game consoles? We conclude with a report on player experiences and showcase some 3D reconstructions built by players during gameplay.

    • 11:45-12:05: Placing a Value on Aesthetics in Online Casual Games, Erik Andersen and Yun-En Liu(PDF Slides)

      Game designers frequently invest in aesthetic improvements such as music, sound effects, animations, and static visuals when polishing their games. However, their exact value for attracting and retaining players remains unclear. Seeking to estimate this value in the domain of online casual games, we conducted a series of large-scale A/B tests in which we selectively removed aesthetic components from two popular Flash games hosted on Kongregate.com. We measured the effect on the play time, progress, and return rate of thousands of players.

      Surprisingly, we found no significant effects of removing music and sound effects from either game. Removing animations caused moderate negative effects. In contrast, making even a small change to gameplay by removing optional rewards dramatically and unexpectedly affected player behavior. Our results suggest that online casual game designers should focus heavily on gameplay, and that audio is not particularly important.

    • 12:05-12:25: Snapshots from inside the CSE Animation Capstone, Barbara Mones

      Our CSE Animation Capstone series of courses culminates in an annual digital short film based on a story written by a UW student. Our students begin the Capstone with no experience in animation or filmmaking. The capstone class experience is designed to be both collaborative and interdisciplinary. We will bring you inside the studio to give you an inside view of our most recent products and the process we use to create them. We will also focus on several innovative approaches we've developed in house to improve the student experience and our production pipeline here at UW.

Session III

    Software Development (CSE 305)

    • 2:40-2:45: Introduction and Overview: David Notkin

    • 2:45-3:05: Deterministic Process Groups in dOS, Tom Bergan(PDF Slides)

      Current multiprocessor systems execute parallel and concurrent software nondeterministically: even when given precisely the same input, two executions of the same program may produce different output. This severely complicates debugging, testing, and automatic replication for fault-tolerance. Previous efforts to address this issue have focused primarily on record and replay, but making execution actually deterministic would address the problem at the root.

      Our goals in this work are twofold: (1) to provide fully deterministic execution of arbitrary, unmodified, multithreaded programs as an OS service; and (2) to make all sources of intentional nondeterminism, such as network I/O, be explicit and controllable. To this end we propose a new OS abstraction, the Deterministic Process Group (DPG). All communication between threads and processes internal to a DPG happens deterministically, including implicit communication via shared-memory accesses, as well as communication via OS channels such as pipes, signals, and the filesystem. To deal with fundamentally nondeterministic external events, our abstraction includes the shim layer, a programmable interface that interposes on all interaction between a DPG and the external world, making determinism useful even for reactive applications.

      We implemented the DPG abstraction as an extension to Linux and demonstrate its benefits with three uses: plain deterministic execution; replicated execution; and record and replay by logging just external input. We evaluated our implementation on both parallel and reactive workloads, including Apache, Chromium, and PARSEC.

    • 3:05-3:25: Weaving Code Extensions into JavaScript, Ben Lerner (PDF Slides)

      Web sites and web browsers have become platforms for delivering sophisticated programs, mostly as JavaScript source. This format has sparked extremely enthusiastic efforts to customize both sites and browsers in ways their authors never expected or accommodated. These customizations take the form of yet more JavaScript dynamically injected into the program, (ab)using JS idioms to patch the running code. This talk presents a language extension to JS to simplify the extension process and remove the need for arcane idioms. Our approach borrows mechanisms from aspect-oriented programming, is efficiently implementable in modern JavaScript compilers, and yields cleaner semantics and better performance than existing idioms.

    • 3:25-3:45: Proactive Detection of Collaboration Conflicts, Mike Ernst (PDF Slides)

      Most software is built by multiple people. Each developer makes decisions about when to incorporate other team members' changes, and when to share changes with other team members. A mistake in these decisions — performing these tasks too early or too late — can lead to delays in development, wasted work, or re-work.

    • This talk presents Crystal, a tool that proactively alerts a developer to: inconsistencies with teammates' work and opportunities to synchronize without suffering conflicts. First, we show that developers make suboptimal choices about when to synchronize their work by analyzing the histories of eight collaborative, real-world, open-source projects. Crystal could let developers know about conflicts, on average, 10 days earlier than they are being resolved today. Crystal could also improve developer confidence that conflict-free changes can be merged safely, on average, 11 days earlier. Second, we perform a systematic study of the space of information that can assist developers in making better-informed decisions, identify and classify the pertinent information, and design a Crystal interface that allows developers to quickly recognize projects and relationships with collaborators that likely required attention. We have deployed Crystal to a handful of developers and used their feedback to improve both the interface and our understanding of collaborative development interactions.

    Computers that Read (CSE 403)

    • 2:40-2:45: Introduction and Overview: Luke Zettlemoyer

    • 2:45-3:05: Learning 5000 Relational Extractors, Raphael Hoffmann(PDF Slides)

      Many researchers are trying to use information extraction (IE) to create large-scale knowledge bases from natural language text on the Web. However, the primary approach (supervised learning of relation-specific extractors) requires manually-labeled training data for each relation and doesn’t scale to the thousands of relations encoded in Web text. In this talk, I will present LUCHS, a self-supervised, relation-specific IE system which learns 5025 relations — more than an order of magnitude greater than any previous approach — with an average F1 score of 61%. Crucial to LUCHS's performance is an automated system for dynamic lexicon learning, which allows it to learn accurately from heuristically-generated training data, which is often noisy and sparse.

    • 3:05-3:25: Mapping Open Information Extraction to a Customer Ontology, Stephen Soderland(PDF Slides)

      Information Extraction (IE) can identify a set of relations from free text to support applications such as Question Answering. Until recently, IE systems were domain-specific and needed a combination of manual engineering and supervised learning to adapt to each target domain. A new paradigm, Open IE operates on large text corpora without any manual tagging of relations, and indeed without any pre-specified relations. Due to its open-domain and open-relation nature, Open IE is purely textual and is unable to relate the surface forms to an ontology, if known in advance. We explore the steps needed to map Open IE tuples to a domain-specific ontology and demonstrate our approach of mapping domain-independent tuples to an ontology using domains from DARPA's Machine Reading Project. Our system achieves precision over 0.90 from as few as 8 training examples for domains of NFL-scoring and Intelligence-Community use cases.

    • 3:25-3:45: Markov Logic in Machine Reading, Hoifung Poon(PDF Slides)

      A long-standing goal of AI and natural language processing is to harness human knowledge by automatically understanding texts. Known as machine reading, it has become increasingly urgent with the rise of billions of web documents. To represent the acquired knowledge that is complex and heterogeneous, we need first-order logic. To handle the inherent uncertainty and ambiguity in extracting and reasoning with knowledge, we need probability. Combining the two has led to rapid progress in the emerging field of statistical relational learning. In this talk, I will show that statistical relational learning offers promising solutions for machine reading. I will present Markov logic, which is the leading unifying framework for representing and reasoning with complex and uncertain knowledge, and has spawned a number of successful applications for machine reading. In particular, I will present USP, an end-to-end machine reading system that can read text, extract knowledge and answer questions, all without any training examples. To resolve variations for the same meaning, USP recursively clusters expressions that are composed with or by similar expressions. In a machine reading experiment, USP extracted five times as many correct answers compared to previous state of the art such as TextRunner, and raised accuracy from below 60% to 91%.

    Battery-free and Low-Power Wireless Sensing (CSE 691)

    • 2:40-2:45: Introduction and Overview: Joshua Smith

    • 2:45-3:05: SNUPI: Sensor Nodes Utilizing Powerline Infrastructure, Ultra-Low-Power, General-Purpose Wireless Sensing, Gabe Cohn(PDF Slides)

      SNUPI nodes are ultra-low-power, general-purpose, 27 MHz wireless sensor nodes that transmit their data by coupling over the powerline to a single receiver attached to the powerline in the home. We’ve demonstrated the ability of our general purpose wireless sensor nodes to provide whole-home coverage while consuming less than 1 mW of power when transmitting (one order of magnitude lower than existing nodes), and our custom CMOS transmitter consumes only 65 µW (two orders of magnitude lower than existing nodes). Compared to those found in traditional whole-home wireless systems, this is the lowest power transmitter to date.

    • 3:05-3:25: WISP & WARP: Far Field Wirelessly Powered Sensing, Alanson Sample

      Radio Frequency signals provide a near ubiquitous energy source due to the large number of TV, cellular, and WiFi transmitters that proliferate our urban environments. While the traditional use of RF transmission is for data transfer, it is possible to harvest, convert, and store this energy for use in a variety of applications. This talk presents ongoing work at Intel Labs Seattle and the University of Washington on RF wireless power harvesting from both ambient and deliberate sources. The goal is to enable new applications and devices that are completely wirelessly powered and have life times that are unconstrained by batteries.

      The Wireless Ambient Radio Power (WARP) project harvests energy from VHF or UHF TV towers. Experimental results show the ability to power a commercially available thermometer/hygrometer, with LCD display, as well as sensor nodes equipped with microcontroller and bidirectional radios, from a range of 4km. While WARP relies on ambient sources of wireless power, the Wireless Identification and Sensing Platform (WISP) makes use of power deliberately transmitted by RFID readers. WISP is a programmable, battery-free sensing and computational platform designed to explore sensor-enhanced RFID applications. WISPs are powered exclusively from RF energy and have an operating range of up to several meters from a standard UHF RFID reader.

    • 3:25-3:45: WREL: Near Field Wireless Power, Joshua Smith

      WREL, the Wireless Resonant Energy Link, is a system for transmitting large amounts of power (tens of watts) over medium distances (several feet). The key feature that differentiates WREL from other wireless power transfer schemes is that it uses a control system to maintain constant efficiency even as operating range and orientation vary. After explaining the technology, I will discuss future directions and potential applications.

Session IV

    HCI (CSE 305)

    • 3:55-4:00: Introduction and Overview: James Fogarty

    • 4:00-4:20: OneBusAway, Brian Ferris(PDF Slides)

      We present our on-going work with OneBusAway (http://onebusaway.org), our award-winning suite of tools to help improve the usability of public transit for over 40k Puget Sound transit riders every week. Public transit is an important tool for those looking to ease their commutes, reduce their car dependence, or perhaps minimize their environmental impact. Unfortunately, the usability of transit systems often leaves much to be desired. OneBusAway attempts to adress that issue by providing a variety of transit information tools for area bus riders across a number of interfaces, including web, phone, SMS, and mobile devices. We describe our current system and along with future directions for the program.

    • 4:20-4:40: Gestalt: An IDE for Machine Learning, Kayur Patel

      We present Gestalt, a development environment designed to support the process of applying machine learning. While traditional programming environments focus on source code, we explicitly support both code and data. Gestalt allows developers to implement a classification pipeline, analyze data as it moves through that pipeline, and easily transition between implementation and analysis. Our experiments shows that Gestalt significantly improves the ability of developers to find and fix bugs in machine learning systems.

    • 4:40-5:00: Prefab: Modifying Any Graphical User Interface, Morgan Dixon

      We propose a vision in which anybody can modify any interface of any application. For example, an HCI researcher developing a new interaction technique might implement and evaluate it in several real-world applications (e.g., Adobe Photoshop, Apple iTunes, Microsoft Office). A practitioner or hobbyist who sees the researcher's publication might then add the technique to several of their favorite applications.

      To explore this vision, we propose pixel-based interpretation to enable modification of interfaces without their source code and independent of their underlying toolkit implementation. I will describe our preliminary work on pixel-based methods in the Prefab system, which reverse engineers the interface structure of graphical applications in real time using only their pixels. I will describe how Prefab can be used to modify any interface of any application.

    Configurable Computing (CSE 403)

    • 3:55-4:00: Introduction and Overview: Carl Ebeling

    • 4:00-4:20: Hardware Acceleration of the Short Read Genome Reassembly Problem, Maria Kim(PDF Slides)

      DNA sequencing determines the nucleotide bases of a genome, which enables us to discern the functionality of organisms. In recent years, new technologies have increased the speed of DNA sequencing significantly. This "Next Generation Sequencing" process takes a strand of DNA and slices it into short DNA chunks to be sequenced in parallel, greatly accelerating the process. However, there are various constraints to reassembling the DNA pieces back into its original form, which is a critical step in determining the information stored in DNA. This reassembly step can, in fact, slow down the entire process of DNA sequencing. Our algorithm takes advantage of FPGA technology, and uses hardware to accelerate the process of reassembly.

    • 4:20-4:40: A High-Level Programming Model for Large-Scale Reconfigurable Computing Platforms, Corey Olson(PDF Slides)

      Large-scale massively parallel computing platforms will continue to push the Exascale computing barrier, but power limitations will require the use of accelerator technologies, such as FPGAs. Specialized, fine-grained computational modules can be loaded onto an FPGA, thereby gaining power efficiency and speed due to the massively parallel nature of an FPGA architecture. However, the design of these FPGA systems requires long design cycle times and a vast knowledge of hardware design. We propose a parallel programming model for applications targeting these large-scale reconfigurable architectures, which abstracts the communication and hardware design from the programmer, thereby enabling them to focus on application development. We will develop a solution to the short read re-assembly problem as an example of this programming model.

    • 4:40-5:00: Efficient Compilation of Complex Data-Dependent Code to Coarse-Grained Reconfigurable Architectures, Stephen Friedman(PDF Slides)

      In reconfigurable architectures, complex data-dependent control flow is often implemented using if-conversion and loop-flattening to create predicated code. This results in allocating hardware resources for all code paths, even though the results of many of those code paths will be discarded. In this talk, I will give an overview of our efforts to efficiently support this type of control flow by sharing resources across mutually exclusive code paths in reconfigurable architectures.

    Sustainability Sensing (CSE 691)

    • 3:55-4:00: Introduction and Overview: Shwetak Patel

    • 4:00-4:40: Closing the Gap Between Action and Effect: Using Computers to Sense and Feedback Resource Consumption Information in the Home, Jon Froehlich

      Why is it that despite spending a fraction of the time in our cars compared to our homes, most of us know much more about the price of gas and the efficiency of our vehicles than we do about the cost of electricity and the energy efficiency of our homes? In this talk, I will introduce our recent work on sensing electricity, gas, and water usage in the home and show how with simple, user-installable sensors, we can provide itemized information to the homeowner about the consumption of specific electrical, gas, and water appliances and fixtures. Our ultimate goal is to inform and empower a wide variety of stakeholders from homeowners and building operators to utilities and policy makers with an unparalleled level of data on resource consumption. Our work promises to transform the ways in which we think about our own homes.

    • 4:40-5:00: WATTR: Self-Powered Water Sensing in the Home, Tim Campbell

      The WATTR device is a novel self-powered sensor that uses changes in a home's water pressure as both a powering and sensing source. WATTR is capable of sampling home water pressure and wirelessly transmitting when any water fixture in the home is opened or closed. WATTR provides an alternative sensing solution to the power intensive Bluetooth-based sensor used in the HydroSense project for single-point whole-home water usage. Unlike other water-based power harvesters, WATTR does not waste water to power itself because it harvests energy from changes in pressure not flow. Finally, WATTR is a viable self-powered sensor capable of monitoring and transmitting water usage data without the use of a battery.