Early Career Researcher Symposium 2018 – Poster Presentations
Poster presenters at the 2018 Early Career Researcher Symposium included early career faculty members, post-docs, and graduate students from many fields of computer science. 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.
Research done by the NetSys lab at Stony Brook.
In this project, a paradigm-shifting anatomy learning system based on screen-based augmented reality techniques (REFLECT or Magic Mirror) is proposed and validated with a large-scale study among premedical students (n=288) at Johns Hopkins University. The results indicated that using REFLECT enhanced students’ engagement, and knowledge retention in the human anatomy with elevated performance in evaluation tests, high level of engagement and increased time to focus on the painting task.
This poster outlines three primary research project that I have been focused on that are related to different uses of robots for law enforcement applications. The first project is Therabot – A Robot Therapeutic Support System. Mental health disorders are a prominent problem across the world. An effective treatment has been the use of animal-assisted therapy; however not everyone can interact with and/or care for a live animal. The second project involves the use of robots for gathering sensitive information from children related to eyewitness memory and bullying victimization. This forms the foundation we hope to transform the way law enforcement can gather sensitive information from children. The third project is the use of robots integrated with tactical teams to improve performance, scene understanding, and situation awareness for better decisions and safety of officers and civilians.
Networked information technologies, such as the internet of things, wearable devices, ubiquitous sensing, and social sharing platforms increase the flow of rich, but often personal, information available for computing research. The growth in the scale, scope, speed, and depth of human data research requires reconsideration of ethical assumptions. This poster introduces PERVADE, an NSF-funded multi-institution project conducting empirical research on Pervasive Data Ethics.
Visual Blends are a communication technique in journalism, ads and PSAs. We show that this difficult and creative task can be decomposed and distribution to both crowds of people and machines to collaboratively create images to communicate important messages.
Omar Chowdhury
The University of Iowa
LTEInspector: A Systematic Approach for Adversarial Testing of 4G LTE
In this poster, we present our findings of analyzing the security and privacy of the three critical procedures of the 4G LTE protocol (i.e., attach, detach, and paging). Our findings include potential design flaws of the protocol and unsafe practices employed by the stakeholders. For exposing vulnerabilities, we propose a model-based testing approach, dubbed LTEInspector, which lazily combines a symbolic model checker and a cryptographic protocol verifier. Using LTEInspector, we have uncovered 10 new attacks along with 9 prior attacks, categorized into three abstract classes (i.e., security, user privacy, and disruption of service), in the three procedures of 4G LTE. Notable among our findings is the authentication relay attack that enables an adversary to spoof the location of a legitimate user to the core network without possessing appropriate credentials. To ensure that the exposed attacks pose real threats and are indeed realizable in practice, we have validated 8 of the 10 new attacks and their accompanying adversarial assumptions through experimentation in a real testbed. Here, we propose a computational approach that integrates genomic and epigenomic data to prioritize patients at risk of treatment resistance. We have integrated DNA methylation and mRNA expression patient profiles, which defined a comprehensive panel of markers of therapeutic response.
Kenneth Fletcher
University of Massachusetts Boston
Collaborative Learning using LSTM-RNN for Personalized Recommendation
Today, the ability to track users’ sequence of online activities, makes identifying their evolving preferences for recommendation practicable. Despite the availability of such information, most existing time-based recommender systems focus on predicting some user rating. This work considers the rich, user activity sequence, and combine the concept of collaborative filtering with long short-term recurrent neural network (LSTM-RNN), to make personalized recommendations. Specifically, the encoder-decoder LSTM-RNN is employed to make sequence-to-sequence recommendations.
Detecting and correcting errors in computer code, also known as debugging, is a fundamental skill for computer programmers. However, explicit and deliberate teaching of this skill is often overlooked in introductory programming courses. Peer Evaluations is an activity designed to help students practice their debugging skills by exposing them to hundreds of faulty programs written by their peers. This activity was implemented in a large introductory programming course with promising results.
The poster presents an optimization framework for bandwidth optimal delivery of data/service to connected vehicles via edge devices.
Vivan Genaro Motti
George Mason University
Assisting Students with Intellectual Disabilities in Inclusive Education with a Smartwatch Application
Smartwatches have a large potential to assist students with disabilities in their everyday activities. However, their potential as assistive technologies in inclusive academic environments is unclear. To investigate how smartwatches can support students with intellectual and developmental disabilities (IDDs) to perform activities that require emotional and behavioral skills and involve communication, collaboration and planning, we implemented WELI. WELI (Wearable Life) is a wearable application designed to assist young adults with IDDs attending a postsecondary education program. Through a user-centric design process, this project investigates how smartwatches can effectively assist students with IDDs in special education. The results reported are drawn from 8 user studies with 58 participants in total. WELI features include behavioral intervention, mood regulation, reminders, checklists, surveys and rewards. Results indicate that several considerations must be taken into account when designing for students with IDDs. Overall, the study participants were enthusiastic with an innovative smartwatch application to be used in class, and reacted positively about the technology and features provided.
There are many settings where there is a desire to disseminate sensitive data as long as privacy is protected. Differential privacy (DP) has emerged as an important standard for privacy-preserving computation, offering rigorous protection while still enabling accurate data analysis. However, deploying DP currently requires a team of privacy experts. This poster describes barriers to real-world deployment and new systems that will make it easier for non-experts to use and deploy DP technology.
Antonina Mitrofanova
Rutgers University
Integrative (epi) genomic analysis to predict response to androgen-deprivation therapy in prostate cancer
Therapeutic resistance is an emerging clinical problem, with detrimental implications in oncology. Here, we propose a computational approach that integrates genomic and epigenomic data to prioritize patients at risk of treatment resistance. We have integrated DNA methylation and mRNA expression patient profiles, which defined a comprehensive panel of markers of therapeutic response. We have demonstrated that this panel predicts patients with predisposition to resistance and those who would benefit from the therapy. Even though driven by a critical need to investigate resistance to androgen-deprivation therapy in prostate cancer, our approach is applicable to a wide range of therapeutic regimens.
Lung cancer is the leading cause of cancer-related deaths in the past several years. A major challenge in lung cancer screening is the detection of lung nodules from computed tomography (CT) scans. State-of-the-art approaches in automated lung nodule classification use deep convolutional neural networks (CNNs). However, these networks require a large number of training samples to generalize well. This paper investigates the use of capsule networks (CapsNets) as an alternative to CNNs. We show that CapsNets significantly outperforms CNNs when the number of training samples is small. To increase the computational efficiency, our paper proposes a consistent dynamic routing mechanism that results in 3 times speedup of CapsNet. Finally, we show that the original image reconstruction method of CapNets performs poorly on lung nodule data. We propose an efficient alternative, called convolutional decoder, that yields lower reconstruction error and higher classification accuracy.
Ifeoma Nwogu
Rochester Institute of Technology
Computational Social Dynamics: Analyzing the Face-level Interactions in a Group
We present a model to measure interactional synchrony (IS) based on facial action units. IS refers to how the speech or behavior of people in a conversation becomes more synchronized, and they can appear almost to mimic one another. The model is composed of intermediary long short-term memory networks, useful for learning the extent of IS between two or more dynamic signals. On a real-life dataset the model successfully estimated the extent of IS, with an overall prediction mean of 2.96% error, as compared to 26.1% random permutations serving as the control baseline.
My work explores roles AI can play in the videogame development process: e.g., visualizing the possibility space of play, verifying that game levels can be beaten, and understanding interactive software in ways legible to humans. As an example, I show how leveraging connections between action games and cyber-physical systems can lead to new insights for hybrid systems identification, general game playing, modeling languages and verification, and interactive software development workflows.
The Wellness Technology lab examines how novel software tools can empower communities to overcome barriers to wellness. We build and evaluate software that helps families learn from personal health data, systems that help people to advocate for and enact change that makes wellness more achievable in their neighborhoods, and tools that help people overcome wellness barriers faced in particular life stages, such as caring for a sick loved one.
The topic of this talk lies in the area of adaptive data analysis, where the goal is to design mechanisms that can give statistically valid answers to adaptively generated queries. My poster presents fast mechanisms for answering adaptive queries into datasets. These mechanisms provide exponential speed-ups per query without increasing sample complexity. These techniques also yield improved bounds for adaptively optimizing convex and strongly convex functions over a dataset.
Wei (Wilbur) Shi
Arizona State University
Decentralized Optimization over Networks with Geometric Convergence
In this poster, we discuss solving decentralized convex optimization problems in a network, where each node has its own convex cost function and the goal is to minimize the sum of the nodes’ cost functions while obeying the network connectivity structure. We present a push-protocol based method for which we can prove its geometric convergence over time-varying networks with simplex communication links. We then present a push-pull-protocol based method and show that the new method converges at a geometric rate over time-invariant networks but illustrates a superior performance over time-varying networks in our numerical experiments. The push-pull based method also unifies algorithms with different types of distributed architecture, including decentralized (peer-to-peer), centralized (master-slave), and semi-centralized (leader-follower) architecture.
Contemporary software development has transitioned from a solo and mainly technical activity to a highly social and collaborative endeavor. The rise and growth of Global Software Engineering and Open Source Software (OSS) development stimulated this increased collaboration. Many developers now contribute to multiple projects, desiring to engage with, learn from, and co-create with other developers, and thereby forming communities of practice. The next generation of software developers needs experience in real-life collaborative software development settings, which will prepare them with more than technical skills. Leveraging the rich variety of OSS projects, new developers can face and eventually overcome barriers to joining these collaborative projects. Additionally, the OSS repositories provide rich information that can be mined and analyzed by techniques such as machine learning and natural language processing. In this seminar, I will focus on my research on understanding the onboarding behavior of newcomer developers and how it laid the foundation to provide support for them. I will introduce the quantitative and qualitative methodological approaches I apply, and then explain in depth how I modeled the barriers and proposed a portal to support newcomers’ onboarding. I then show how this study inspired other research lines and future opportunities to support social developers in this new landscape.
Chenhao Tan
University of Colorado Boulder
Friendships, Rivalries, and Trysts: Characterizing Relations between Ideas in Texts
Understanding how ideas relate to each other is a fundamental question in many domains, ranging from intellectual history to public communication. Because ideas are naturally embedded in texts, we propose the first framework to systematically characterize the relations between ideas based on their occurrence in a corpus of documents, independent of how these ideas are represented. Combining two statistics — cooccurrence within documents and prevalence correlation over time — our approach reveals a number of different ways in which ideas can cooperate and compete. For instance, two ideas can closely track each other’s prevalence over time, and yet rarely cooccur, almost like a “cold war” scenario. We observe that pairwise cooccurrence and prevalence correlation exhibit different distributions. We further demonstrate that our approach is able to uncover intriguing relations between ideas through in-depth case studies on news articles and research papers.
Tammy Toscos
Parkview Health
“Give Me More Consumer Health Technology”: A Multi-Platform Intervention to Support Medication Adherence
This presentation will showcase a novel, multi-platform intervention that was designed using a patient centered approach and deployed in a controlled 6-month trail aimed at improving adherence to oral anti-coagulant therapy in patients living with atrial fibrillation. Participants will learn how various consumer health technologies were integrated into the intervention that uses a novel algorithm for delivering tailored messaging via Epic’s patient portal, MyChart. Prevention of thromboembolism (stroke) is key in individuals with non-valvular atrial fibrillation (AF) and nonadherence to oral anticoagulant (OAC) therapy is a frequent causative factor of stroke. There are many factors that contribute to the complexity of medication non-adherence among the AF patient population, including patient knowledge about AF and related therapies, the patient’s health beliefs and concerns, patient-provider communication, particular reminder systems and habits that patients develop to help them consistently and correctly take medication. Thus, interventions aimed at changing this behavior must be tailored to address the unique internal motivations and information needs of those confronted with the daily choice to take a medication to prevent stroke.
Da Yan
The University of Alabama at Birmingham
T-thinker: A Task-Centric Framework to Revolutionize Big Data Systems Research
Existing Big Data systems are designed for data-intensive problems where data movement is the bottleneck and CPUs are underutilized. However, many real-life problems have a high computational complexity, such as community detection and training random forest. This poster introduces a new task-centric framework, T-thinker, for truly scalable distributed compute-intensive Big Data processing. As a preliminary work, T-thinker has been applied to graph mining: https://info.cs.uab.edu/yanda/gthinker/
Lana Yarosh
University of Minnesota
Technology for Empowering Social Connections in Critical Contexts
Social isolation is a critical challenge for mental health and wellbeing. People are self-identifying as “lonely” at higher rates than ever and many are experiencing effects of social isolation, including depression, and substance use disorders. My work takes on the task of designing technologies to increase connectedness. I work closely with people in specific critical social contexts to design and evaluate the effects of new technologies on social connectedness. The poster elaborates on my work in two critical contexts: parenting & intergenerational mentorship and peer support in health.
Yunpeng Zhang
University of Houston
Interactive Based Access Control Framework for Connected Vehicle Communication
Threats of attack on Vehicle Communications pose the risk of serious and unexpected consequences. We proposes a novel Software Defined Networking based Global Access Control (SGAC) to accommodate Vehicle communication systems. SGAC enforces strict rules and dynamic changes based on network activities to yield better consistency and provide information at different levels of abstractions to enable the smooth flow of traffic. The simulation results showed that the SGAC could effectively control the mean speed throughout the entire work zone area.
These were recorded at the Computing Community Consortium (CCC) Symposium for Early Career Researchers, which was supported by the National Science Foundation under Grant No. 1136996. 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.
