Colloquium

Colloquium talks will be held in the Katharine Reed Cudahy Building, Room 401 on the Marquette University campus. Please address inquiries/suggestions to Dr. Michael Zimmer.

Spring 2020 - Unless specified, the talks will begin at 1:00pm CT in Room 414 at Cudahy Hall.

  • Tuesday, January 14: Dr. Maia Jacobs (Center for Research on Computation and Society, Harvard)
    Title: One Size Doesn’t Fit Anyone: Tailoring Digital Tools for Personal Health Journeys
    Personal technologies for everyday health management have the potential to transform healthcare by empowering individuals to engage in their own care, scaffolding access to critical information, and supporting patient-centered decision-making. Currently, many personal health tools often focus only on a single task or isolated event. However, chronic illnesses are characterized by information needs and challenges that shift over time; thus, these illnesses are better defined as a dynamic trajectory than a series of singular events.
     
    In this talk, I discuss my work designing and implementing novel computing systems that: 1) support chronic illness trajectories and 2) reduce patients’ barriers to health information access. I’ll present my approach using personalized and adaptive content to connect individuals with timely and actionable feedback. Using results from longitudinal field deployments, I demonstrate the ability for these tools to facilitate patients’ proactive health management and engagement in their care. I’ll also discuss the utility of this approach for encouraging to long-term engagement with health tools, as evidenced by longitudinal usage patterns. I’ll conclude with opportunities for using personalization as a strategy to support other complex information tasks, including the health management of illness trajectories in which uncertainty is paramount and the integration of machine learning models into clinical workflows.
     
  • Wednesday, January 15: Saeed Tizpaz Niari (CS Department, CU Boulder) 
    Title: Differential Debugging for Side Channel Vulnerabilities
    In early 2018, Meltdown and Spectre attacks challenged the security of any computer devices globally. These attacks exploit timing information to compromise the confidential information of users. While most existing debugging techniques provide supports for the functional correctness, the support for non-functional properties such as information leaks via timing observations is scarce. In this talk, Tizpaz-Niari will showcase a range of tools and techniques to detect, explain, and mitigate side-channel vulnerabilities in large-scale libraries and web applications. The technique combines tools from gray-box fuzzing, dynamic program analysis, and machine learning inferences. The talk also presents a novel technique to adapt neural networks for quantifying the amounts of information leaks.
     
  • Thursday, January 16: Dr. Kevin Moran (Computer Science, William & Mary) 
    Title: Toward Practical Automation for Software Engineering
    Given the ubiquity of software in modern society and its applications in increasingly complex problem domains, today’s developers require practical automation in order to effectively and efficiently build, test, and maintain software systems. At the same time, the proliferation of software has led to the creation of an unprecedented amount of freely available data that describes a diverse array of software systems. Artifacts such as source code files, screenshots, videos and bug reports provide a wealth of information from which patterns can be learned to enable useful automation for developers. In this talk, I will describe two of my recent research projects that use machine learning techniques to harness this data contained within software repositories to automate different components of the development lifecycle for mobile applications. My focus on mobile apps stems from the potential impact given the popularity of mobile platforms among both developers and users coupled with development challenges that are unique to the mobile domain (such as change-prone APIs and platform fragmentation). These projects aim to improve developer productivity while alleviating the effects of these challenges.
     
    First, I will introduce an approach that completely automates the process of prototyping of GUIs for mobile apps. This approach, called ReDraw, is able to transform an image of a mobile app GUI into runnable code by detecting discrete GUI-components using computer vision techniques, classifying these components into proper functional categories (e.g., button, dropdown menu) using a Convolutional Neural Network (CNN), and assembling these components into realistic code. Second, I will present a technique that is capable of translating a screen-recording of a mobile application into a replayable scenario. This technique, called V2S, is based primarily on computer vision techniques and adapts recent solutions for neural object detection and image classification to detect and classify user actions on a screen. Finally, I will conclude my talk by providing a brief overview of related projects, as well as my research vision and planned future work.

Fall 2019 - Unless specified, the talks will begin at 1:00pm CT in Room 414 at Cudahy Hall.

  • November 25 - Mustafa Bilgic, Department of Computer Science, Illinois Institute of Technology. Title: Active and Interpretable Classification                                                                             Abstract: I will talk about active and interpretable classification research we do at the Machine Learning Laboratory at Illinois Institute of Technology in Chicago. I will primarily focus on active learning, learning with rationales, interpretable machine learning, and active inference. Application domains include text classification, anomolous flight detection, and wireless sensor networks. If time permits, I will also summarize our recent work on relational classification of cells in microscopy images.                                                                                                                                             Bio: Dr. Mustafa Bilgic is an Associate Professor, the Director of the Masters in Artificial Intelligence, the Co-Director of the BS in Artificial Intelligence, and the Director of the Machine Learning Laboratory in the Computer Science Department at Illinois Institute of Technology. He received his BS in Computer Science from the University of Texas at Austin and his MS and PhD in Computer Science from the University of Maryland at College Park. He received an NSF CAREER award and his current research is supported by NSF and Samsung. His research interests include active and interpretable machine learning, statistical relational learning, and probabilistic graphical models.
  • November 18 - Elena Zheleva, Department of Computer Science, University of Illinois, Chicago Title: Causal inference and counterfactual learning from relational data
    Abstract: Predictive systems are used to make sense of rapidly increasing amounts of data and support human decision making, from what to read, to whom to date, whether to invite a job applicant for an interview, and what drug to develop next. However pervasive, these systems are limited in their ability to reason about cause and effect, and the consequences of their own behavior. In this talk, I will motivate the need for developing algorithms that can answer causal questions from real-world data which is inherently noisy, relational, and multimodal. I will focus on the problem of heterogeneous treatment effect estimation, understanding the individual variations in outcomes between individuals. I will show how learning algorithms can address this problem and take into account people's individual characteristics and predispositions. I will demonstrate the value of these algorithms in the context of understanding what drives empathy online and whether a given piece of information will go viral.
  • November 4 - Bilge Mutlu, Department of Computer Sciences, University of Wisconsin Madison     Title: Designing Robots for Human Interaction
    Abstract: Robots are emerging as a family of technologies that offer their users unprecedented capabilities through highly complex, situated, and continuous interactions, reshaping many human activities and environments. These new capabilities and forms of interaction also bring forth new design challenges—how do we build such technologies so that they work effectively with people, that they are not disruptive human social and physical environments, and that they are accepted, adopted, and welcomed by people? In this talk, I will present research that addresses three key challenges of designing robotic technologies for human interaction: (1) defining and systematically studying the design space for robotic systems for human use; (2) developing and using design support tools for the complex design problems introduced by these systems; and (3) integrating robotic systems into human environments and studying their use and adoption over long periods of time.
  • October 30 - Sameer Patil, School of Informatics, Computing and Engineering, Indiana University, Bloomington.                                                                                                                                   Title: Mental Models and User Experiences of the Tor Browser
    Abstract: With the exponential increase in government and corporate surveillance of online activities, there is an increasingly important need for usable tools that help individuals maintain privacy. While the Tor Browser is a popular anonymity tool, it has yet to achieve notable levels of mainstream usage by non-expert users. Making the Tor Browser appealing to the general population would require greater attention to usability and user experience aspects. To this end, we carried out two studies to examine user understanding of Tor operation and user experience of browsing the Web using the Tor Browser, respectively. The first study found significant differences in the mental models of experts and non-experts regarding Tor operation and threat model. The second study uncovered a number of significant challenges users encounter when using the Tor Browser for everyday online activities. Based on these findings, we offer a number of suggestions for making the Tor Browser more usable, thus helping boost privacy and anonymity for everyone.
  • October 21 - Marina Kogan, School of Computing, University of Utah                                        Title: Human-Centered Data Science for Crisis Informatics
    Abstract: Natural disasters are associated with the disruption of existing social structures, but they also result in the creation of new social ties by those affected as they problem-solve alone and together. With social media now being a site for some of this interaction, there is much to learn about the nature of those changing social structures, including how and why they shift. However, the study of this social arena is challenging, because the high-tempo, high-volume convergent nature of crisis events produces vast amounts of social media data, necessitating the use of the data science methods. On the other hand, to glean meaningful insight from the crisis-related social media activity, it is necessary to use methods that account for the complex social context of the user activity. In this talk Kogan will show how the Human-Centered Data Science provides methodological approaches that both harness the power of computational methods and account for the highly situated nature of social media activity in disruption. Utilizing these methodological approaches, she will show how disaster-related coordination and distributed problem-solving take shape on two social media platforms: Twitter and OpenStreetMap.
  • October 14 - Stevie Chancellor, Department of Computer Science, Northwestern University.   Title: Human-Centered Machine Learning for High-Risk Health Behaviors OnlineSocial networks are now adopting machine learning-based approaches to identify and remove unwanted behaviors and content that harm online communities. One example of these behaviors is high-risk mental health behaviors, such as suicidal ideation, pro-eating disorder, and opioid addiction – these behaviors are particularly dangerous because they can have contagion-like effects on others, and the communities that support them evade platform moderation and governance. There is an urgent need to innovate data-driven and automated systems to handle content like this, yet these approaches often oversimplify the context and complexities of mental disorders and unique effects communities have on people and on platforms. In short, these high-risk mental health behaviors are a rich domain area to explore tensions in applying machine learning to complex social behaviors in online communities.

    In this talk, I will describe how I use human-centered machine learning to solve problems in this space. I’ll talk about the interdisciplinary intersection of machine learning with other disciplines like clinical psychology and critical data studies to make rigorous and ethical predictions of people’s behaviors. I will talk about my work in applied Machine Learning and Computational Linguistics enable robust computational models that identify mental health signals in social media. I will then discuss recent research on ethics and research practices, and how this informs our understanding of responsible and compassionate algorithm design. I’ll conclude with how human-centered insights can be brought to other domains to leverage computational methods to answer our toughest questions about deviant behavior online.
  • October 11 - Blair Taylor, Department of Computer and Information Sciences, Towson University.
  • October 7 - Arjun Krishnan, Department of Computational Mathematics, Science and Engineering, Michigan State University.                                                                                         Title: Hurdles in the marathon of data-driven biology
    Abstract: Big data collections contain valuable signals that can help fill critical gaps in our biomedical knowledge. Our group is focused on developing computational approaches to leverage these data and build predictive models that link genes to various aspects of health and disease. This data-driven approach is, though, fraught with challenges from the get go. For instance, close to 1.5 million human gene-expression profiles (along with another million from other species) are publicly available. However, seamlessly using all these data for integrative analysis is not straightforward because (a) majority of publicly-available samples lack complete basic information about source such as tissue, sex, age, disease status, and environmental conditions and (b) depending on the technology/platform, different sample captured the expression of different subsets of genes in the genome, leading to thousands of missing/unmeasured genes in many of these profiles. In this talk, I will talk about these and other challenges and our approaches for addressing them.
  • September 16 - Mike Slattery, Department of Computer Science, Marquette University            Title: Quantum Computing for Beginners, or, How I Spent My Summer
    Abstract: I plan to talk about the IBM initiative "IBM Q Experience" which allows anyone to access some quantum computers.  I'm going to explain how to think about computing with these machines and how to actually do it.  The talk will be a bit more technical than a typical colloquium talk, but will only assume the audience has a general idea of current digital circuits (AND gate, OR gate, etc.).

    Part 2: We'll look at some of the algorithms described on IBM's site

  • September 9 - Mike Slattery, Department of Computer Science, Marquette University              Title: Quantum Computing for Beginners, or, How I Spent My Summer
    Abstract: I plan to talk about the IBM initiative "IBM Q Experience" which allows anyone to access some quantum computers.  I'm going to explain how to think about computing with these machines and how to actually do it.  The talk will be a bit more technical than a typical colloquium talk, but will only assume the audience has a general idea of current digital circuits (AND gate, OR gate, etc.).

    Part 1: We'll look at the theory behind quantum computing and how to run quantum circuits using IBM's Circuit Composer.

Spring 2019

Colloquium dates and speakers for Spring 2019 - Unless specified, the talks will begin at 2:00pm CT in Room 401 at Cudahy Hall.

Fall 2018

Spring 2018

Fall 2017

Spring 2017