Computing Research Symposium – 2016 Poster Presentations
Poster presenters at the CCC Symposium on Computing Research included early career faculty members, post-docs, and graduate students from many fields of computer science. There were a total of 39 poster presentations during the poster reception. The videos below are from the poster presenters who chose to record a short video clip of their presentation.
To see the full list of poster presenters at the Symposium and their abstracts, please see the CCC Symposium Poster Book.
Matthew Bietz
University of California, Irvine
Personal Data for Public Health: Opportunities and Challenges
New forms of data can help us understand human health. Streams of data and traces of human behavior are being generated by mobile apps, wearable devices, online activity, social media and the Internet of Things. My research investigates the opportunities and challenges for using these new forms of personal data for research in public health, including issues of privacy, research ethics, and the potential biases in these datasets.
Yonathan Bisk
University of Southern California
Natural Language Communication with Computers
We propose a framework for devising testable algorithms for bridging the communication gap between humans and robots. We begin with a setting in which humans give instructions to robots using unrestricted natural language commands, with instruction sequences aimed at building complex goal configurations in a blocks world. We then flip the paradigm to also address the problem of language generation, where a human performs commands produced by a machine to demonstrate two-way communication.
Gedare Bloom
Howard University
Security and Provenance in the Internet of Things
Embedded and cyber-physical systems are increasingly vulnerable to attack especially because of ubiquitous network access: the Internet of Things (IoT) refers to the proliferation of such systems. The biggest challenge that IoT brings is the multifaceted attack surfaces spanning embedded systems, mobile applications, cloudservices, and the gateways and networks connecting them. This project aims to improve device security using low-cost solutions in open source embedded system software for IoT end points.
Bistra Dilkina
Georgia Tech
Computational Sustainability @ GT
We present some of the work in Computational Sustainability done at Georgia Tech. In particular, we focus on large-scale combinatorial optimization models and methods to facilitate cost-effective biodiversity and landscape connectivity conservation. With increasing habitat fragmentation and shifts due to urban growth, climate change and sea level rise, it is key to develop effective data-driven and scalable decision support tools that can help policy makers take informed choices.
Vivian Genaro Motti
George Mason University
Wearable Health: Exploring Human-Centered Solutions of On-Body Technologies to Improve Healthcare
Wearable technologies have a large potential to support healthcare. Their solutions not only span across different stages and domains of medical care but also benefit care givers and patients. By understanding the potential opportunities that emerge from wearables in healthcare from a human centered perspective, we aim at bringing technologies seamlessly in to users’ lives, supporting their daily activities and bridging the gap between what users need and what wearable technologies provide them.
Christopher Harris
Brown University
Processor Design Exploration for Vision Based Mobile Robots
Vision based mobile robots such as UAVs require increasingly complex onboard processors. These systems require high-performance, low power computing be integrated with actuators and data-intensive sensors in a system providing real-time performance. This research describes a new modeling framework for the design of specialized system-on-chip (SoC) modules for sensor based robots. This work represents a new and untapped frontier at the intersection of robotics and computer architecture.
Katie Henry
Johns Hopkins University
Automated In-patient Monitoring in the ICU with Application to Septic Shock Prediction
Clinicians are continually monitoring which patients in the ICU are at high of risk of developing an adverse event. We can leverage machine learning to develop models that use higher dimensional input to identify developing events earlier than traditional approaches, which typically rely on paper-based scores. In particular we demonstrate the ability to automatically identify septic shock, a leading cause of death in the United States, many hours earlier than current clinical practices.
Bert Huang
Virginia Tech
Weakly Supervised Cyberbullying Detection in Social Media
A growing majority of human communication occurs online. Advances in mobile and connected technology amplify individuals’ abilities to interact. Unfortunately, the amplification of social connectivity includes a disproportionate amplification of detrimental behavior such as cyberbullying. We are developing automated, social-aware, and data-driven methods for cyberbullying detection. These methods could enable technologies that mitigate the harm and toxicity created by these harmful behaviors.
Patrick Gage Kelley
University of New Mexico
Privacy as Iconography: A Pictographic Collection and Comparison
The promise of icons as a silver bullet for simplifying the complexity of privacy policies, permissions, data sharing, and other similarly tangled privacy applications remains unmet. This poster explores the visual language of these icons, across platforms and contexts, how they have evolved over the last decade, and attempts to intuit what the future of privacy icons holds.
Bart P. Knijnenburg
Clemson University
Adaptive Privacy Decision Support
Privacy concerns are an important barrier to the growth of user data-driven applications. To help users balance the benefits and risks of information disclosure in a user-friendly manner, I developed a “privacy adaptation procedure” that offers tailored privacy decision support. This procedure predicts users’ privacy preferences and behaviors based on their past behavior and known characteristics. It then provides automatic adaptive default settings in line with users’ disclosure profile.
Parisa Kordjamshidi
University of Illinois Urbana-Champaign
Declarative Learning Based Programming
Developing intelligent problem solving systems for real world applications is very challenging. I present Saul, a declarative machine-learning based programming language that is designed to help experts in various domains who are not expert in machine learning, to design complex data-driven intelligent systems. Such a language helps reusability and replicability of research results and facilitates the of use structured learning learning algorithms, new data resources and background knowledge.
Su-In Lee
University of Washington
Big Data Approach to Identify Novel Biomarker in Cancer
Cancer is fully of mysteries. Two individuals with seemingly similar tumors sometimes have very different responses to chemotherapy and other treatments, as well as drastically different survival outcomes. In order to better understand this phenomenon, researchers have developed ways to obtain a molecular snapshot of an individual’s tumor, which amounts to gigabytes of data. I will present our recent development of novel machine learning approaches to identify biomarkers from these big data.
Xiaohui Liang
University of Massachusetts Boston
Towards Privacy-preserving Data Release in Mobile Healthcare
Mobile healthcare (mHealth) integrates advanced sensing and communication technologies to continuously monitor the data that affects and reflects the private health status of patients. The privacy-preserving data release in mHealth is challenging because the release process over a huge volume of data must be efficient, fast, and accurate. In this poster, we will present new query-based data release and differentially private data release approaches to address the unique challenges of mHealth.
Jonathan May
University of Southern California
Advance in Rapid-Response Low-Resource Machine Translation
In 2010, response to the devastating earthquake in Haiti was hampered by an inability of international aid workers to understand Haitian Creole. Machine Translation (MT) has helped bridge the language barrier, however, state-of-the-art quality is generally limited to cases where hundreds of millions of words of human translation have already been collected. I will show how recent advances in neural network learning can be leveraged to yield state-of-the-art MT results with limited resources.
Mahsa Mirzargar
University of Utah
Empowering Uncertainty Characterization Using Scientific Visualization
Data-intensive methods are transforming scientific discovery. Visualization, as an integral component of the data analysis, can facilitate the exploration and communication of the data and the uncertainty. In this poster, I will present novel uncertainty visualization paradigms that use nonparametric statistical analysis to derive robust summaries from uncertain data. I will demonstrate their utility for analyzing and visualizing simulation outputs used to predict the path hurricanes will take.
Pavithra Prabhakar
Kansas State University
Robust Verification of Cyber-Physical Systems
The poster describes the recent results and software tools for automated analysis of robustness of cyber-physical systems, with particular focus on the interaction of discrete and continuous dynamics that arise due to the interaction of digital controllers with physical systems. The techniques borrow ideas from formal methods, dynamical systems theory and control theory.
Mohamed Sarwat
Arizona State University
GeoExpo-Interactive and Scalable Exploration of Big GeoSpatial Data
Recently, the volume of spatial data increased tremendously. Such data includes but not limited to: weather maps, socioeconomic data, and vegetation indices. Making sense of spatial data is beneficial for several applications that may transform science and society, e.g., socio-economic, climate change analysis, urban planning, road network design, and fast disaster response. This poster presents GeoExpo a data management system that enables interactive and scalable exploration of spatial data.
Anshumali Shrivastava
Rice University
Scalable, Sustainable and Secure Machine Learning via Probabilistic Hashing
Modern big-data settings pose a new set of challenges which primarily include scalability, energy efficiency and privacy. I design probabilistic hashing algorithms which provide a practical and provable solutions to all the three major challenges. These algorithms trade a very small amount of certainty, with often exponential gains in the computations, memory and energy efficiency. They also provide secure probabilistic encoding of the data which do not reveal the attributes information.
Ting Wang
Lehigh University
Serendipity or Preparedness? Quantifying Creativity in Scientific Enterprise
The building blocks of scientific innovations are often embodied in existing knowledge, yet it is creativity that blends disparate ideas. Here, by correlating researchers’ information consumption with their publications, we find remarkable predictability in scientific creative processes. Further, we develop a mechanistic model that not only effectively predict disparate ideas likely to be linked by creativity, but also identifies critical references for such linkings to happen.
Dong Wang
University of Notre Dame
Data Reliability in Cyber-Physical Systems for Smart Cities
Consider a Cyber-Physical System (CPS) application that uses crowdsensing to collect data about the physical environment. The data reliability challenge in CPS refers to designing a state estimator that takes raw unreliable crowdsensing data as input and outputs reliable estimates of the underlying physical states and appropriate error bounds on estimations. This poster presents interesting perspectives of addressing this emerging challenge and its connection to future smart city applications.
Hongning Wang
University of Virginia
Online Collaborative Learning from Interaction with Humans
It is vital for an online service system to learn from its interactions with humans. In our work, we have developed bandit-based algorithms to enable a system to 1) actively learn from its interaction with a crowd of users; 2) identify latent factors that facilitate online learning. We rigorously prove nice theoretical properties of the proposed online learning algorithms, and verified their utilities with extensive experiments in various real-world applications.
Gloria Washington
Howard University
Affective Technologies for Improving the Lives of Persons with Chronic Disease
Behavioral and physical characteristics can be used as input to technologies that monitor stress, anxiety, and/or depression in humans. These technologies are commonly found in academic laboratories and not where we need them: mobile devices. This work seeks to improve the lives of sickle cell disease sufferers through motivational, encouraging technologies that also provide dietary tips. Results from this work will help to alleviate the emotional toll for managing and handling chronic disease.
Ting Zhu
University of Maryland, Baltimore County
Gait-Based Privacy Preserving Technique for Smart Cities
With the advent of the Internet of Things (IoT) and big data, high fidelity localization and tracking systems that employ cameras, RFIDs, and attached sensors intrude on personal privacy. However, the benefit of localization information is important for smart city applications. To address this challenge, we introduce Wobly, an attribute based signature (ABS) which measures gait. Wobly uses the physical layer channel and the unique human gait as a means of encoding a person’s identity.
These were recorded at the Computing Community Consortium (CCC) Symposium on Computing Research, which was supported by the National Science Foundation under Grant No. 1019343. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.
