(Knowledge) Graph Data Analytics

Visual Graph Data Analysis
Unlike traditional relational database management systems (RDBMS), graph databases (GDBs, e.g., Neo4j, TigerGraph, DGraph, ArangoDB) store data together with its accompanying relationships. This holistic data understanding allows us to model non-trivial, e.g., multi-relational or (in)direct deep relationships, such as the ones occurring in multi-omics data, fraud analysis, or sustainable supply chains. The VIG group is concerned with building better and visually driven graph analytics approaches targeting the challenge of Multivariate Network Visualization/Analysis, not only for the fraud detection use case.

For example, within the GraphPolaris project we develop a no-code analytics platform for graph analysis. It enables non-data scientists to contribute their domain knowledge and analyze large and complex datasets without the typically required query scripting and allows exposing the gathered analytical insights directly through compelling visualizations.

For this project, we are actively searching for BSc/MSc students, who are interested in research and development. If you are interested contact Michael Behrisch (m.behrisch AT uu.nl)

Funding: This work is partially sponsored by NWO and industry sources.

VR4eVR: Virtual Reality for enhanced Visual Rehabilitation

In this project, several teams (clinical centers, serious game developers, and academic institutions) combine their multidisciplinary skills to create and deploy a solution for rehabilitating patients with visual impairments by offering them training scenarios using a virtual reality (VR) setup, and by offering a visual analytics (VA) dashboard to clinicians to monitor patient progress.

Funding: This project is financed by the Dutch Research Council (NWO) research programme Knowledge and Innovation Covenant (KIC).

Information Visualization

Information visualization is used to depict and explore non-spatial data in a visual way using good visual metaphors and efficient interaction techniques. My research focuses on large multivariate graph and trail-set simplified visualization via bundling, image-based techniques for non-spatial data, and high-dimensional data visualization and via dimensionality reduction. To approach all these challenges, I design new scalable InfoVis techniques and tools using GPU programming. At a higher level, I am working on theory, techniques, and tools to unify scientific visualization and information visualization via imaging and computer graphics.

Software Visualization and Data Mining

Software visualization methods help understanding and maintaining large and complex code bases by presenting the structure, attributes, and evolution of source code in scalable and intuitive ways. We develop methods that extract various types of information from source code, program traces, and software repositories: dependencies, structure, quality metrics, and developer activity. Next, we develop ways to show the structural evolution of code at class, function, or statement level and that combine the visualization of software architecture diagrams with software metrics defined on groups of diagram elements. We implement our methods in tools that are used on real-world software systems.

Multiscale Shape Processing

Multiscale shape processing involves 2D and 3D shape segmentation, denoising, edge and feature detection, simplification, and matching done on several levels of detail. These operations can be effectively and efficiently supported by several classes of methods, such as skeletonization (or medial axis computation), level sets, and partial differential equations. Applications include image and volume inpainting, surface reconstruction from point clouds, image and volume compression, shape segmentation, and shape denoising and simplification.

Visual Analytics

Visual analytics aims at generating and proving hypotheses to support understanding of complex phenomena based on large data sets. Visual analytics is tightly connected with data science, data mining, and machine learning. My research focuses on supporting machine learning engineering by explaining the operation of complex engines such as classifiers and providing interactive tools to incorporate user insights into their design and optimization processes. At a high level, my aim is to empower both machine learning engineers with tools that aid them to design and fine-tune their engines, and end-users of these tools in understanding how and why machine learning took its decisions.

Surveys and State-of-the-Art Reports

State-of-the-art reports present the major developments over one or more decades in a research subfield. They highlight the development of the area, best results achieved, provide a taxonomy to organize existing work, and highlight hot active directions. Surveys quantitatively and/or qualitatively compare methods, techniques, and tools to help practitioners in choosing optimal ones for a given context. I have worked on several surveys related to the above-described areas of my research.