CSE Doctoral Student Seminar: Son Dinh and Shali Jianghttps://engineering.wustl.edu/Events/Pages/CSE-Doctoral-Student-Seminar-Son-Dinh-and-Shali-Jiang.aspx832CSE Doctoral Student Seminar: Son Dinh and Shali Jiang2017-03-24T05:00:00Z12:30 p.m.2 p.m.Lopata Hall, Room 101<p><strong>"Scheduling Platforms and Techniques for Parallel Soft Real-time Systems​"</strong></p><p><strong>Son Dinh </strong><br/>Adviser: Chris Gill</p><p>Multicore processors are becoming common nowadays. In order to take the advantage of multicore processors, programs need to be parallelized. Real-time applications are no exception from this trend, especially when real-time applications with high computation demand are emerging. However, the question of how to efficiently schedule parallel real-time tasks on multicore processors is still an open question. In this talk, we will experimentally investigate two common strategies for scheduling parallel applications, namely centralized scheduling and randomized work stealing. We also examine these scheduling approaches for soft real-time tasks when combining with federated scheduling, a real-time scheduling paradigm for parallel tasks which theoretically guarantees parallel tasks to meet their timing constraints (i.e., deadlines) when running on their own dedicated cores. <br/></p><p><strong><strong>"Efficient Nonmyopic Active Search"</strong><br/></strong></p><p><strong>Shali Jiang</strong><br/>Adviser: Roman Garnett<br/></p>Active search is an active learning setting with the goal of identifying as many members of a given class as possible under a labeling budget. In this work, we first establish a theoretical hardness of active search, proving that no polynomial-time policy can achieve a constant factor approximation ratio with respect to the expected utility of the optimal policy. We also propose a novel, computationally efficient active search policy achieving exceptional performance on several real-world tasks. Our policy is nonmyopic, always considering the entire remaining search budget. It also automatically and dynamically balances exploration and exploitation consistent with the remaining budget, without relying on a parameter to control this tradeoff. We conduct experiments on diverse datasets from several domains: drug discovery, materials science, and a citation network. Our efficient nonmyopic policy recovers significantly more valuable points with the same budget than several alternatives from the literature, including myopic approximations to the optimal policy.<p><br/></p>
CSE Colloquia Series: David Jurgenshttps://engineering.wustl.edu/Events/Pages/CSE-Colloquia-Series-David-Jurgens.aspx834CSE Colloquia Series: David Jurgens2017-03-24T05:00:00Z11 a.m.Lopata Hall, Room 101<p style="text-align: center;"><strong>People in Context: Social Understanding through Linguistic and Network Analysis</strong></p><p style="text-align: center;"><strong>David Jurgens</strong></p><p style="text-align: center;">Postdoctoral Scholar</p><p style="text-align: center;">Department of Computer Science</p><p style="text-align: center;">Stanford University</p><p><strong>Abstract</strong></p><p>The rise of new online platforms for capturing many aspects of our daily lives has opened the door for large-scale computational studies of nearly all facets of behavior, such as exercise, collaboration, and community dynamics.  My research aims to understand and predict these kinds of human behaviors by combining techniques from natural language processing and network science to produce holistic models of people and their social interactions.  The first part of this talk focuses on the challenge of inferring the demographics that describe who people are.  Using the example of location inference, I show how we can efficiently and accurately learn these aspects for hundreds of millions of people across the globe.  I also discuss my recent work on algorithmic bias in demographic inference and show how to mitigate inequality for the ubiquitous task of language identification.  The second part of the talk shifts from individuals to their social interactions and I describe my recent work analyzing how people's offline behavior and communication strategies change when they join online groups.  I conclude by highlighting future directions in computational social science that I am excited to pursue through the combined lens of language and networks.</p><p><strong>Biography</strong></p><p>David Jurgens is postdoctoral scholar in the department of Computer Science at Stanford University and received his PhD from UCLA.  His research combines natural language processing, network science, and data science to discover, explain, and predict human behavior in large social systems. He was recently awarded a Volkswagen Foundation grant for his work on population modeling to measure the influence of international actors on national news topics.  He is currently the Data Co-Chair for ICWSM and his research on demographic inference has been featured in news outlets such as the MIT Technology Review, Forbes, and Business Insider.<br/></p><p><br/></p>
CSE Colloquia Series: Yufei Dinghttps://engineering.wustl.edu/Events/Pages/CSE-Colloquia-Series-Yufei-Ding.aspx843CSE Colloquia Series: Yufei Ding2017-03-27T05:00:00Z10:30 a.m.Jolley Hall, Room 309<p style="text-align: center;"><strong>High-Level Program Optimizations for Data Analytics</strong></p><p style="text-align: center;"><strong>Yufei Ding </strong><strong> </strong></p><p style="text-align: center;">Ph.D. Candidate</p><p style="text-align: center;">Computer Science Department</p><p style="text-align: center;">North Carolina State University</p><p><strong>Abstract</strong></p><p>Many modern applications, especially those data analytics, often spend a large number of cycles on unnecessary computations. To find a document most similar to a query document, for instance, these applications typically would need to examine hundreds of thousands of other documents (that are not the most similar ones) in the dataset. Such redundant computations have been hidden in the useful instructions of the applications and are elusive for traditional compiler-based code optimizations. My work harnesses these hidden but significant optimization opportunities by raising the level of program optimizations from implementations to algorithms, and from instructions to formulas. <br/></p><p><strong>Biography</strong></p><p>Yufei Ding is a Ph.D. candidate in the Computer Science Department at North Carolina State University. She received her B.S. and M.S. in Physics from University of Science and Technology of China and The College of William and Mary respectively. In 2012, she started her Ph.D. study in Computer Science.  Her research interest resides at the intersection of Compiler Technology and (Big) Data Analytics, with a focus on enabling High-Level Program Optimizations for data analytics and other data-intensive applications. Yufei has been actively publishing in major venues in both computer systems and data analytics areas, such as ASPLOS, PLDI, VLDB, ICDE, and ICML. She was the receipt of NCSU Computer Science Outstanding Research Award in 2016.<br/></p>
CSE Colloquia Series: Mohit Iyyerhttps://engineering.wustl.edu/Events/Pages/CSE-Colloquia-Series-Mohit-Iyyer.aspx844CSE Colloquia Series: Mohit Iyyer2017-03-31T05:00:00Z11 a.m.Lopata Hall, Room 101<p style="text-align: center;"><strong>Using Deep Learning to Understand Creative Language</strong></p><p style="text-align: center;"><strong>Mohit Iyyer </strong><strong> </strong></p><p style="text-align: center;">Ph.D. Candidate</p><p style="text-align: center;">Department of Computer Science</p><p style="text-align: center;">University of Maryland, College Park</p><p><strong>Abstract</strong></p><p>Creative language—the sort found in novels, film, and comics—contains a wide range of linguistic phenomena, from phrasal and sentential syntactic complexity to high-level discourse structures such as narrative and character arcs. In this talk, I explore how we can use deep learning to understand, generate, and answer questions about creative language. I begin by presenting deep neural network models for two tasks involving creative language understanding: 1) modeling dynamic relationships between fictional characters in novels, for which our models achieve higher interpretability and accuracy than existing work; and 2) predicting dialogue and artwork from comic book panels, in which we demonstrate that even state-of-the-art deep models struggle on problems that require commonsense reasoning. Next, I introduce deep models that outperform all but the best human players on quiz bowl, a trivia game that contains many questions about creative language. Shifting to ongoing work, I describe a neural language generation method that disentangles the content of a novel (i.e., the information or story it conveys) from the style in which it is written. Finally, I conclude by integrating my work on deep learning, creative language, and question answering into a future research plan to build conversational agents that are both engaging and useful.</p><p><strong>Biography</strong></p><p>Mohit Iyyer is a fifth year Ph.D. student in the Department of Computer Science at the University of Maryland, College Park, advised by Jordan Boyd-Graber and Hal Daumé III. His research interests lie at the intersection of deep learning and natural language processing. More specifically, he focuses on designing deep neural networks for both traditional NLP tasks (e.g., question answering, sentiment analysis) and new problems that involve understanding creative language. He has interned at MetaMind and Microsoft Research, and his research has won a best paper award at NAACL 2016 and a best demonstration award at NIPS 2015.<br/></p>
CSE Colloquia Series: Daniel Larremorehttps://engineering.wustl.edu/Events/Pages/CSE-Colloquia-Series-Daniel-Larremore.aspx884CSE Colloquia Series: Daniel Larremore2017-04-03T05:00:00Z10:30 a.m.Jolley Hall, Room 309<p style="text-align: center;"><strong>Stochastic Generative Models for Complex Networks</strong></p><p style="text-align: center;"><strong>Daniel Larremore </strong></p><p style="text-align: center;">Omidyar Postdoctoral Fellow</p><p style="text-align: center;">Santa Fe Institute</p><p><strong>Abstract</strong></p><p>Understanding real-world network datasets requires tools to identify patterns and methods to quantify whether those patterns are noteworthy and meaningful. Stochastic generative models satisfy both these needs in mathematically principled ways by specifying parameters of a stochastic data-generating process that results in an ensemble of networks, each with an associated probability of having been generated by the process. When we engineer the probabilities to be uniform over the ensemble, we can treat the generative model as a null model to measure whether an empirically observed network property is normal or surprising. When we instead use parameters to engineer the probabilities to be biased toward particular types of networks—for example, those with community structures, groups, or clusters—we can infer the parameters that best explain empirical data, thereby detecting communities, extracting hierarchies, or identifying correlations in the process. The properties of the generative model impact not just the types of structures that can be identified, but also the efficiency with which ensemble parameters can be inferred from real data and how rapidly the ensemble can be sampled. Therefore, careful mathematical choices about generative models can drastically improve our ability to make useful predictions and can also enable us to rigorously analyze the performance of statistical inference of network structure. I will introduce stochastic generative models in the context of two applied problems: the evolution of malaria parasite virulence genes and the movement of scholars in the academic labor market. In the process, these investigations will reveal provable limits to the detection of community structures in complex networks which apply beyond the framework of stochastic generative models to network science more broadly.</p><p><strong>Biography</strong></p><p>Daniel Larremore is an Omidyar Fellow at the Santa Fe Institute. His research develops statistical and inferential methods for analyzing large-scale network data, and uses those methods to solve applied problems in diverse domains, including public health and academic labor markets. Prior to joining the Santa Fe Institute he was a post-doctoral fellow at the Harvard T.H. Chan School of Public Health 2012-2015. He obtained his Ph.D. in Applied Mathematics from the University of Colorado at Boulder in 2012, and holds an undergraduate degree from Washington University in St. Louis.<br/></p>
CSE Colloquia Series: Alice Gaohttps://engineering.wustl.edu/Events/Pages/CSE-Colloquia-Series-Alice-Gao.aspx875CSE Colloquia Series: Alice Gao2017-04-07T05:00:00Z11 a.m.Lopata Hall, Room 101<p style="text-align: center;"><strong>Eliciting and Aggregating High-Quality Information from the Crowd</strong><br/></p><p style="text-align: center;"><strong>Alice Gao </strong><strong> </strong></p><p style="text-align: center;">Postdoctoral Fellow</p><p style="text-align: center;">Department of Computer Science</p><p style="text-align: center;">University of British Columbia</p><p><strong>Abstract</strong></p><p>Accurate information is essential for solving many problems.  Such information often exists as dispersed knowledge and beliefs of many people.  I am interested in solving problems by collecting and aggregating a large amount of information contributed by many self-interested individuals.  My goal is to motivate participants to provide high-quality information and to aggregate their reports to inform decision making.</p><p>In this talk, I will describe two pieces of my work.  First, I will introduce prediction markets, which are mechanisms for aggregating information for forecasting future events.  Prediction markets outperformed alternative forecasting methods in various settings, but they failed catastrophically at predicting the outcomes of Brexit and the 2016 US presidential election.  My work considers scenarios in which prediction markets may fail and offers insights on possible reasons for such failures.</p><p>Next, I will tackle the problem of grading assignments in large classes.  One way to provide timely feedback to students is peer grading --- having students evaluate one another.  A challenge is to motivate students to invest sufficient effort in providing accurate evaluations.  I design mechanisms, which provides incentives for effort and accuracy while using limited ground-truth evaluations provided by teaching assistants.  </p><p>Finally, I will outline my future research agenda on designing effective mechanisms for eliciting and aggregating dispersed information.</p><p><strong>Biography</strong></p><p>Alice Gao is a postdoctoral fellow in Computer Science at the University of British Columbia, advised by Kevin Leyton-Brown. She is generously supported by the Canadian NSERC Postdoctoral Fellowship.  Alice's research is on designing mechanisms for eliciting and aggregating dispersed information.  Her work has tackled a range of problems including forecasting future events and grading assignments in large classes.  Alice obtained her Ph.D. in Computer Science from Harvard University advised by Yiling Chen and her Bachelor's degree in Computer Science and Mathematics from the University of British Columbia.<br/></p>
CSE Colloquia Series: Chien-Ju Hohttps://engineering.wustl.edu/Events/Pages/CSE-Colloquia-Series-Chien-Ju-Ho.aspx861CSE Colloquia Series: Chien-Ju Ho2017-04-10T05:00:00Z10:30 a.m.Jolley Hall, Room 309<p style="text-align: center;"><strong>Learning and Incentives in Systems with Humans in the Loop</strong></p><p style="text-align: center;"><strong>Chien-Ju Ho </strong></p><p style="text-align: center;">Postdoctoral Associate</p><p style="text-align: center;">Cornell University</p><p><strong>Abstract</strong></p><p>There is an increasing amount of human-generated data available on the internet -- including online reviews, user search histories, datasets labeled using crowdsourcing, and beyond. This has created an unprecedented opportunity for researchers in machine learning and data science to address a wide range of problems. On the other hand, human-generated data also creates unique challenges. Humans might be strategic or careless, possess diverse skills, or have behavioral biases. What is the right way to understand and utilize human-generated data? Furthermore, can we better design the systems with humans in the loop to generate more useful data in the first place?</p><p>In this talk, I will present my research which addresses the challenges in utilizing and eliciting data from humans. In particular, I will introduce the problem of actively purchasing data from humans for solving machine learning tasks, and demonstrate how to convert a large class of machine learning algorithms into pricing and learning mechanisms. I will also discuss how to obtain high-quality data from humans using financial incentives and present our findings in a comprehensive set of behavioral experiments conducted on Amazon Mechanical Turk.</p><p><strong>Biography</strong></p><p>Chien-Ju is a postdoctoral associate at Cornell University.  He obtained his Ph.D. in Computer Science from UCLA in 2015, advised by Jenn Wortman Vaughan. He also spent three years visiting the EconCS group at Harvard from 2012 to 2015, hosted by Yiling Chen. His research interests are in machine learning, algorithmic economics, online behavioral social science, crowdsourcing, and artificial intelligence. His dissertation was on the design and analysis of crowdsourcing mechanisms. He is the recipient of the Google Outstanding Graduate Research Award at UCLA in 2015. His work was nominated for Best Paper Award at WWW 2015.<br/></p>
CSE Colloquia Series: Ulugbek Kamilovhttps://engineering.wustl.edu/Events/Pages/CSE-Colloquia-Series-Ulugbek-Kamilov.aspx892CSE Colloquia Series: Ulugbek Kamilov2017-04-12T05:00:00Z10:30 a.m.Jolley Hall, Room 309<p style="text-align: center;">​<strong style="text-align: center;">SEAGLE: Robust Computational Imaging under Multiple Scattering</strong></p><p style="text-align: center;"><strong>Ulugbek S. Kamilov </strong><strong> </strong><strong> </strong></p><p style="text-align: center;">Research Scientist</p><p style="text-align: center;">Computational Sensing Team</p><p style="text-align: center;">Mitsubishi Electric Research Laboratories </p><p><strong>Abstract</strong></p><p>Majority of modern methods in high-resolution three-dimensional (3D) optical microscopy rely on linear scattering models that assume weakly scattering samples, making them inherently inaccurate for many applications. This places fundamental limits—in terms of resolution, penetration, and quality—on the imaging systems relying on such models. In this talk, we describe a new technique for computational imaging called SEAGLE that combines a nonlinear scattering model and a total variation (TV) regularized inversion algorithm. SEAGLE exploits an efficient representation of light scattering as a recursive neural network for formulating a fast, large-scale imaging algorithm. The key benefit of SEAGLE is its efficiency and stability, even for objects with large permittivity contrasts. SEAGLE is suitable for robust imaging under multiple scattering and has a potential to broadly impact 3D imaging of multicellular organisms such as biological tissue.</p><p><strong>Biography</strong></p><p>Ulugbek S. Kamilov is a Research Scientist in the Computational Sensing team at Mitsubishi Electric Research Laboratories (MERL), Cambridge, MA, USA. Dr. Kamilov obtained his B.Sc. and M.Sc. in Communication Systems, and Ph.D. in Electrical Engineering from the École Polytechnique Fédérale de Lausanne (EPFL), Switzerland, in 2008, 2011, and 2015, respectively. In 2007, he was an Exchange Student at Carnegie Mellon University (CMU), Pittsburgh, PA, USA, in 2010, a Visiting Student at Massachusetts Institute of Technology (MIT), Cambridge, MA, USA, and in 2013, a Visiting Student Researcher at Stanford University, Stanford, CA, USA.</p><p>Dr. Kamilov's research focus is computational imaging with an emphasis on the development and analysis of large-scale computational techniques for biomedical and industrial applications. His research interests cover imaging through scattering media, multimodal imaging, optical microscopy, and subsurface imaging. He has co-authored 17 journal and 32 conference publications in these areas. His Ph.D. thesis work on Learning Tomography (LT) was selected as a finalist for EPFL Doctorate Awards 2016 and was featured in the "News and Views" section of the Nature magazine. Since 2016, Dr. Kamilov is a member IEEE Special Interest Group on Computational Imaging.<br/></p>
CSE Colloquia Series-Hyunwoo Kimhttps://engineering.wustl.edu/Events/Pages/CSE-Colloquia-Series-Hyunwoo-Kim.aspx899CSE Colloquia Series-Hyunwoo Kim2017-04-14T05:00:00Z11:00 a.m. Lopata Hall, Room 101<p style="text-align: center;"><strong>Statistical Machine Learning on Manifolds for Structured Data (Without the Pain)</strong></p><p style="text-align: center;"><strong>Hyunwoo J. Kim</strong></p><p style="text-align: center;">Ph.D. Candidate</p><p style="text-align: center;">Department of Computer Sciences<br/></p><p style="text-align: center;">University of Wisconsin-Madison</p><p><strong>Abstract</strong></p><p>Manifold-valued data naturally occur in many disciplines. For example, directional data can be represented as points on the unit sphere. Diffusion tensors in magnetic resonance images form a quotient manifold GL(n)/O(n), which is a space of symmetric positive definite (SPD) matrices. Also, the Hilbert unit sphere can be used for the square-root representation of orientation distribution functions (ODFs) or probability density functions (PDFs). Their data spaces are known, a priori, to have a nice mathematical structure with well-studied properties. It makes sense that if algorithms make use of this additional information, even more efficient inference procedures can be developed. Motivated by this intuition, in this talk we study the relationship between statistical learning algorithms and the geometric structures of data spaces encountered in machine learning, computer vision and neuroimaging using mathematical tools (e.g. Riemannian geometry). As a result, this framework gives new insights into statistical inference methods for image analysis and enables developing new models for manifold-valued data (and potentially manifold-valued parameters) to improve statistical power </p><p><strong>Biography</strong></p><p>Hyunwoo J. Kim is a Ph.D. candidate in the Department of Computer Sciences at University of Wisconsin-Madison (Ph.D. minor: statistics). He earned a B.S. degree and an M.S. in computer science at Korea University and Seoul National University respectively. His research interests include statistical machine learning and manifold statistics for structured data with applications in computer vision and medical imaging. He is actively collaborating with the Wisconsin Alzheimer's Disease Research Center (ADRC) at UW-Madison.<br/></p>
Groundbreaking Ceremony: East Danforth Campus Transformationhttps://engineering.wustl.edu/Events/Pages/Jubel-Hall-Groundbreaking.aspx524Groundbreaking Ceremony: East Danforth Campus Transformation2017-05-05T05:00:00Z4 p.m.<ul><li>Anabeth and John Weil Hall<br/></li><li>Henry A. and Elvira H. Jubel Hall (Engineering)<br/></li><li>Gary M. Sumers Welcome Center<br/></li><li>James M. McKelvey, Sr. Hall (Engineering)<br/></li><li>Mildred Lane Kemper Art Museum Expansion<br/></li><li>Ann and Andrew Tisch Park<br/></li></ul>
Commencement: Engineering Student Recognition Ceremonyhttps://engineering.wustl.edu/Events/Pages/Engineering-Student-Recognition-Ceremony.aspx567Commencement: Engineering Student Recognition Ceremony2017-05-18T05:00:00Z1:30 p.m.Field House, Athletic Complex<p>Students (undergraduate, graduate and PhD) should arrive for lineup before 12:45 p.m. in the lower level hallway. Upon arrival, students should obtain a name card and complete the information requested on the back. <span style="line-height: 25.6px;">Students should carry (not wear) hoods.</span><br/></p><p><strong>Reception locations following the ceremony:</strong></p><ul><li><strong>Biomedical Engineering: </strong>Whitaker Hall</li><li><strong>Computer Science & Engineering:</strong> Sever Plaza</li><li><strong>Electrical & Systems Engineering:</strong> Green Hall </li><li><strong>Energy, Environmental & Chemical Engineering:</strong> Brauer Hall </li><li><strong>Mechanical Engineering & Materials Science:</strong> Lopata Hall</li></ul>Kim Shilling, (314) 935-6100
Commencementhttps://engineering.wustl.edu/Events/Pages/All-University-Commencement-2017.aspx568Commencement2017-05-19T05:00:00Z8:30 a.m.Brookings Quadrangle<p>Engineering students lineup along Louderman Hall. <span style="line-height: inherit;">PhD stude</span><span style="line-height: inherit;">nts assemble with the Graduate School next to Wilson Hall.</span>​</p>