2017 Research Showcase Exhibits
The 70+ exhibits at our Research Showcase will give you a chance to connect with leading researchers from the department who are diving into the major tech areas on the horizon: big data, machine learning, security, robotics, and much more.
*Note: this list is updated as exhibits are registered for the event.
Faculty member or research group:
Extensible Programming Languages: Techniques and Tools
Eric Van Wyk
Extensible programming languages allow users to automatically and reliably add new domain-specific features to their programming language to raise the level of abstraction of their language to that of the task at hand. We demonstrate these capabilities with ableC - an extensible specification of C and language extensions for parallel programming, database access, and other domains.
Constructing Reliable, Safe, and Secure Cyber-Physical Systems
The Critical Systems Group (CriSys) performs research in embedded and safety-critical systems, where software errors could lead to loss of life, environmental damage, or large monetary losses. We aim to create a comprehensive framework for critical software development, focusing on the difficult and often misunderstood areas of software requirements specification and validation/verification.
Spatial Computing Research Group
Spatial Computing is ideas and technologies that transform our lives by understanding the physical world, knowing and communicating our relation to places in that world. The transformational potential of Spatial Computing is already evident. Spatial Computing Research Group focuses on the storage, management and analysis of scientific and geographic data, information and knowledge.
Finding Semantically-Equivalent Binary Code By Synthesizing Adaptors
Independently developed codebases typically contain segments of code that perform the related operations (semantic clones). Such related segments often have different interfaces, so some glue code (an adaptor) is needed to replace one with the other. We present an algorithm that searches for replaceable code segments by attempting to synthesize an adaptor between them from a family of adaptors.
Automatic emulator testing tool made faster.
Bit-Vector Model Counting using Statistical Estimation
My poster will be about approximate model counting for bit-vector SMT formulas (generalizing #SAT) with adding random parity constraints (XOR streamlining). I propose an approach inspired by statistical estimation to continually refine a probabilistic estimate of the model count for a formula, so that each XOR-streamlined query yields as much information as possible.
Retargetable and Behaviorally-Accurate Dynamic Binary Translation (DBT)
Pen-Chung Yew and Antonia Zhai
Dynamic binary translation (DBT) is an enabling technology for system virtualization. In this NSF-sponsored project, we explore the core issues in designing and implementing efficient retargetable DBT for clouds that supports a client/server environment with different vendors’ binaries and platforms. A verification framework is also provided to test the correctness of the dynamic binary translator.
Social Media Analytics
Jaideep Srivastava (Data Mining Research [DMR] Group)
We aim to study social network evolution and information spread over them. We explore higher order networks like Hypergraphs for group evolution prediction and learning group representations. On the social media side, we are interested in leveraging trust networks to learn user representations to classify them as rumor spreaders and also to identify seed users that maximize content virality.
Computational Sleep Science: A Machine Learning Approach
The rapid pace of modern existence has resulted in an increasing prevalence of poor sleep quality, and boosted interest in studying sleep behaviours and their contributing factors. There is a critical and urgent need for robust, automated, high-fidelity algorithms.
High Performance Graph and Tensor Analysis
We develop scalable parallel algorithms for analyzing sparse and irregular graphs and tensors. We present recent work on accelerating triangle counting and truss decomposition for graphs, and various kernels that arise in sparse tensor factorizations. Our resulting software is often orders-of-magnitude faster than competing approaches.
Improving Higher Education with Learning Analytics
Developing tools to support students and learning is a significant task in today’s educational environment. Our initial steps towards enabling such technologies focused on predicting the student’s performance. We can treat it as a "rating prediction" problem while taking into consideration the existing hierarchies, side information and knowledge components of the courses.
Monitoring Dynamics of Water Bodies at Global Scale using Remote Sensing Data
Fresh water, which is only available in inland water bodies is increasingly becoming unevenly distributed across the world, a situation which is posing a threat to human sustainability. My research aims at developing machine learning methods for analyzing spatio-temporal remote sensing data to help create the first global monitoring system of inland water dynamics.
Group-Specific Learning to Personalize Evidence-Based Medicine
Our work aims to use the data available in the electronic health records to improve personalization of patient care. Specifically, we are developing a modeling framework, Group-Specific Learning (GSL), with the ability to enhance clinical modeling by making models increasingly personalized. To give our work an appropriate scope and leverage our expertise, we focus on type-II diabetes mellitus.
Big Data in Climate
Massive amount of data from Earth observing satellites as well as physics-based earth system models offer huge potential for understanding how the Earth's climate and ecosystem have been changing. This poster will highlight various challenges and opportunities they present for both advancing machine learning as well as the science of climate change.
Interactive Visualization Lab
We will be demonstrating a some of the data visualization techniques developed in our research lab. Try out an immersive experience in forestry data, and see how artists and scientists can work together to make beautiful images from weather data.
Perception and Embodiment in Virtual Reality
Our research explores issues related to spatial perception, presence, and self-embodiment in virtual environments. Through the development of novel technologies focused on bringing VR "to life", we seek to expand the potential of VR to more effectively support collaborative design and other applications.
Current Research in Virtual Reality
We will bring a poster, 48" x 36", that highlights selected research from our lab.
A Flexible Distributed Hypergraph Processing System
With the rapid growth of large online social networks, the ability to analyze large-scale social structure and behavior has become critically important. In this poster, we present MESH, a flexible distributed framework for scalable hypergraph processing, which enables different design choices for the key implementation issues of hypergraph representation and partitioning.
Data Placement in a Multi-cloud Environment
Distributed Computing Systems Group (DCSG)
We first present challenges from exploiting cloud storage services of different cloud providers and then two systems that address those challenges.
1) A policy-driven cloud storage system called Wiera that provides an easy way to specify data placement policy.
2) An automated multi-tiered geo-distributed data placement system called TripS that determine the optimized data placement automatically.
Constellation : An Edge-based Framework for Sharing Across Internet of Things Applications
Distributed Computing Systems Group
The current state of the Internet of Things (IoT) leaves immense levels of performance on the table by failing to recognize commonalities among applications and exploiting them for the mutual benefit of whole IoT systems. We propose Constellation, an edge-based query-oriented network abstraction for creating robust, distributed IoT applications.
Nebula: Dispersed Edge Cloud System
Abhishek Chandra & Jon Weissman (DCSG)
We propose Nebula: an edge cloud platform that consists of storage and computational resources that are distributed around the globe.
Nebula provides supports for wide-area data analytics and mobile-cloud computing by allowing data processing to be performed at the edge of the network or close to the data source, and hence reduce the communication bandwidth needed for analysis.
Cross-Technology Communication via PHY-Layer Emulation
We will demonstrate a use case of controlling ZigBee light bulbs through the WiFi/Bluetooth radios on smartphones to showcase our cross-technology communication design (i.e., direct WiFi -> ZigBee, and direct Bluetooth -> ZigBee).
Towards a More Manageable and Resilient Internet via SDN
As our physical world is more intertwined with the cyber world, and as we are increasingly reliant on it, resilience of cyber infrastructure is critical. Managing networks has become an enormous complex and challenging task, and calls for a new paradigm. Our research proposes novel systems and frameworks to provide unified and resilient management control over existing network infrastructures.
Peer Productions with GroupLens
Online Peer Productions have become a popular mean of knowledge production. Well-known examples include Wikipedia, Open Street Maps, etc. Volunteers work together to produce valuable artifacts that are available to the entire internet. In this exhibit, we will present some of our research work to show some interesting problems in those communities.
Konstan / GroupLens
As we celebrate the 20th anniversary of the MovieLens research recommender system, we showcase a set of innovative new results looking at how to improve user experience through voice interaction, increased interactivity, focusing on recommendation freshness, and accounting for user personality.
We designed and created two Massive Open Online Courses (MOOCs) that introduce Recommender System and UI Design to large audiences around the world. We will share our experience of teaching HCI at scale and creating collaborative learning experiences on MOOCs.
Geography, Context, and Trust in the Sharing Economy and Crowdsourcing
Sharing economy and crowdsourcing platforms facilitate interactions that are collaborative and geographically situated in nature. Our research centers around the context in which those interactions happen, ranging from review of the current literature, qualitative analysis of trust determinants, quantitative examination of geographical biases, to field studies on Pokémon GO and Mechanical Turk.
Technology for Empowering Social Connections in Critical Context
Lana Yarosh (GroupLens)
To address issues on increasing social isolation, our research aims to understand the opportunities for technology to amplify and empower social connectedness in critical social context, develop new interface technologies to increase connectedness, and transition this work from academia to real-world use.
ParaBox: Exploiting Parallelism for Virtual Network Functions in Service Chaining
Service Function Chains (SFCs) comprise a sequence of Network Functions that are typically traversed in-order by data flows. Consequently, SFC delay grows linearly with the length of the SFC. Yet, for highly latency sensitive applications, this delay may be unacceptable. We propose ParaBox that dynamically parallelizes the packet processing when possible.
Hunt For The Unique, Stable, Sparse And Fast Feature Learning On Graphs
For the purpose of learning on graphs, we hunt for a graph representation that exhibit certain uniqueness, stability and sparsity properties while also being
amenable to fast computation. This leads to a discovery of family of graph spectral distances, which we prove to possess most of these desired properties. Finally, we show its superior performance on graph classification task.
Demo: Cross-Technology Communication via PHY-Layer Emulation
Cross-Technology Communication (CTC) is an emerging research direction providing a promising solution to the wireless coexistence problem in the ISM bands. We have a demo showcasing how our CTC designs can be implemented on off-the-shelf smartphones for smart light bulbs control.
Channel Coordination via Cross-Technology Communication
This paper presents ECC that uniquely enables explicit channel coordination among hetero- geneities via cross-technology communication (CTC), while maintaining full compatibility to commodity devices. Unlike any implicit coordination designs, ECC generates the guaranteed white space using WiFi CTS, which is then explicitly notified to ZigBee through CTC for immediate use.
Scalable Transaction Management in Cloud Data Systems
Distributed Systems Lab (Anand Tripathi)
This project has developed protocols for scalable transaction management for NoSQL (HBase) and geo-replicated data storage systems, supporting multiple consistency models.
Beehive framework for Graph Data Analytics and Scalable Location-based Services
Distributed Systems Lab (Anand Tripathi)
Beehive programming framework supports a transactional model of parallel computing for large graph problems. We utilize Beehive in building scalable location-based services using a graph-based computing model.
Infrared Tooth Modelling
The exhibit will consist of an experimental setup consisting of a webcam mounted to a ring stand, an extracted tooth mounted on a mini tripod, and a breadboard circuit powering an infrared LED to show the translucency properties of teeth in the infrared, along with a poster explaining how to use the images generated by the webcam to determine the internal structure of the tooth.
Characterizing Complex Material Appearance
This work seeks to learn and predict how people see and interpret the appearance of complex, spatially-varying materials. Past research has studied specific attributes, like color or glossiness, in isolation of the other appearance modifiers. From that foundational research, we seek to develop a computational model matching human judgments of visual similarities between material appearances.
Interactive Visualization Lab
Recent research in the Interactive Visualization Lab
Interactive Visualization Lab
Recent research in the Interactive Visualization Lab
pEVSL: a parallel EigenValue Slicing Library
We present our recent progress on the development of a parallel eigenvalue slicing library for extracting a large number of eigenvalues and their associated eigenvectors of a Hermitian matrix pencil. The target application is extreme-scale electronic structure analysis and planets' normal mode computations.
Multilabel Classification with Group Testing and Codes
The modern mutlilabel classification problems have very large number of classes. However, each instance belongs to only one or few classes. In this work, we propose a novel approach based on group testing to solve such large multilabel classification problems with sparse label vectors. The proposed approach has several advantages theoretically and practically over existing popular methods.
Topological Properties of Neural Network Computations
By viewing neural networks as graphical objects, one may investigate the topological properties of this graph structure. We investigate how the topological structure of semantic information within these networks affects classification decisions and find, among other insights, that the topological signatures of network information are distinct across input classes.
Identifying Animals in Camera Trap Images with Neural Networks & Citizen Scientists on Zooniverse
(Data Science) Human Computer Optimization - School of Physics and Astronomy
Identification of animal species in camera trap images using convolutional neural networks and Citizen Science annotations from Zooniverse.
NoSQL Data Base HBase Performance Optimization on Top of Object Storage
The read performance of HBase is influenced by two factors: 1) HFile compaction, and 2) large amount of RPC between backend storage and HBase RegionServer. To optimize the read performance, new caching policy, moving the SCAN instance from RegionServer down to the storage nodes and designing the new indexing structure might be the possible choices.
Evaluating Host Aware SMR Drives
We are investigating special features of Host-Aware SMR drives by extensive and thorough evaluations, aiming to have insight of the drive characteristics which will guild us design HA-SMR based storage systems. We study the open zone issue, non-sequential zone issue, media cache cleaning efficiency, etc. We are also proposing an H-Buffer design to reduce the default media cache cleaning severity.
Software Defined Storage in the New Cloud with Docker
Recently, the boom in Docker Containers has led to an inevitable transition from traditional virtual machines to such a more light-weighted virtualization technology in cloud.
In this research, we focus on ensuring storage Service Level Objectives(SLOs) in this new cloud environment where people deploy applications via Docker.
The Interactive Robotics and Vision Lab
Demonstrating work on field robots with human-in-the-loop autonomy, and an underwater robot on-the-bench.
Enhancing Underwater Images using Generative Adversarial Networks
We paper propose a method to improve the quality of visual underwater scenes
using Generative Adversarial Networks (GANs), with the goal
of improving input to vision-driven behaviors further down
the autonomy pipeline. Furthermore, we show how recently
proposed methods are able to generate a dataset for the purpose
of such underwater image restoration.
SafeDrive : A Robust Lane Tracking System for Autonomous and Assisted Driving Under Limited Visibility
Junaed Satter, IRV-Lab
SafeDrive is an approach towards robust lane tracking for assisted and autonomous driving, particularly under poor visibility. SafeDrive attempts to improve visual lane detection approaches in drastically degraded visual conditions without relying on additional active sensors other than those found on a smartphone.
Cucheb: A GPU implementation of the filtered Lanczos procedure
This poster describes the software package Cucheb, a GPU implementation
of the filtered Lanczos procedure for the solution of large sparse symmetric
Center for Distributed Robotics
The major objectives of our research are to design, build and demonstrate distributed robotic systems and state of the art algorithms to accomplish a given mission whether it be cancer detection, environmental monitoring, mental health assessment, among others.
Active Convolutional Neural Networks for Cancerous Tissue Recognition
We utilized active learning techniques to capitalize on the power of Convolution Neural Networks for image processing while reducing the amount of data required to achieve favorable results. This is especially applicable to the cancerous tissue recognition task where limited data is available.
Solar Powered UAV
The design and development of solar-powered aerial vehicles for high endurance sensing and monitoring.
Assistive AI for Aging and Memory Loss
Living with memory and sensory deficits caused by aging presents many challenges to patients and caregivers. These challenges, like repetitive questioning or low vision, can severely affect quality of life. We present an artificial agent as a tool to help users cope with or overcome some of these challenges, with many key features like voice activation that make it uniquely suited to the task.
Reciprocation in Repeated Games
We present a simplified environment to study reciprocation. We discuss motivations for reciprocation and show a method to generate reciprocating strategies for a specified environment.
Time Series Shapelets: To Normalize or Not To Normalize
Shapelets are small subsequences of time series that can be used for fast, accurate classification of unlabeled time series. Previous shapelet research has assumed using Z-normalized distances improves accuracy. Using 85 datasets, we show Z-normalization frequently decreases accuracy and that choosing whether to Z-normalize prior to shapelet learning increases shapelet classification accuracy.
Investigating Social Engagement with a Humanoid Robot: An observational assessment
We hypothesize that a naturalistic play-based assessment between children with ASD and a humanoid robot can elicit meaningful individual differences in social engagement. We show our experimental setup with neurotypical children.
Robots in Wildlife
The goal is to develop algorithms and systems in which autonomous robots collect data in order to survey animal habitats in the wild. Equipped with onboard infra-red and standard cameras, they keep track of the count and distribution of all animals in a given area. They also search for radio signals from collared animals and use the signal to localize them.
Vision-Based Yield Estimation for Apple Orchards
In this project, we develop an end to end computer vision system for estimating yield in an apple orchard using images captured from a single camera. Our system is platform independent and does not require any specific lighting conditions. We verify the performance of our algorithms by conducting multiple field trials over the course of two years.
Data Mangement Lab
Mohamed F. Mokbel
University of Minnesota's data management group aims at broadening the focus of database and data management techniques beyond their traditional scope, such as designing systems that take care of spatial and spatio-temporal data. We do both theoretical and systems in areas such as Spatial databases, stream processing, Location-based Services, Big Data, social networks, and crowd sourcing.
IBRelight: Image-based 3D Rendering of Color Appearance for Cultural Heritage
We present ongoing research towards a practical image-based rendering pipeline for cultural heritage applications. The exhibit will include a demonstration of IBRelight, an image-based renderer that supports relighting for inhomogeneous, non-Lambertian surfaces. A poster will give an overview of the methodology as well as ongoing efforts to validate and refine the software.