Open Topics

Machine Learning:

Interpretability of Critical Points in n-Dimensional Surfaces

Visualizations of multi-dimensional surfaces often involve dimensionality reductions such as low-dimensional projection. Such projections lead to a loss of geometry information and create new axes that are difficult to interpret. Cartesian sampling would avoid this problem, but it becomes computationally expensive (exponential in the number of dimensions) as the number of dimensions increases. Additionally, visualizations become more difficult to interpret. The slicing method by Torsney-Weir, with adaptations by Doknic, could address those challenges. However, it is still unclear how interpretable and useful the resulting visualizations are.

Is it possible to make 4D and 5D spaces interpretable? If yes, can it be generalized for higher dimensions? We expect a detailed analysis of critical points ("Kurvendiskussion") for well-known functions such as the Rosenbrock Function. Here, the critical points are already known. The idea is to start with 1D and 2D functions that are easy to visualize and expand to 3D, 4D, and 5D.

The key part of this research is a user study that addresses the following questions:

  • Is it possible to identify the critical points of a function through visual means?
  • How does the slicing method [1] compare to low-dimensional projections and Schlegel diagrams [2] or topological spines [3]?
  • Can Hasse diagrams [4] or Loewner-John ellipsoids [5] (or similar) help us understand multi-dimensional functions, for example, to support sampling? Literature with examples: Lectures on Polytopes by Günter Ziegler.


[1] Sliceplorer
[2] Schlegel diagrams
[3] C. Correa, P. Lindstrom, P.Bremer, Topological Spines: A Structure-preserving Visual Representation of Scalar Fields, see https://ieeexplore.ieee.org/document/6064947
[4] B. A. Davey and H. A. Priestley, Introduction to Lattices and Order
[5] John ellipsoids, see https://arxiv.org/abs/2501.01801

Prerequisites: VIS, FDA
Contact: Torsten Möller, Aleksandar Doknic, Daniel Pahr

Library for Visualization of Slices

This is primarily a programming project. Slicing methods are a novel way of visualizing multi-dimensional data. However, there is no publicly-available library for R or Python that makes it easy to use these visualization techniques. The goal of these projects is to develop such a library. Students should have knowledge of Javascript and either R or Python.

 

Prerequisites: VIS
Programming languages: Javascript and (R or Python)

Contact: Torsten Möller, Thomas Torsney-Weir

Scalable Sampling with Lattices

The challenge is easy to describe:

Come up with an algorithm that takes as input N (the number of samples) and D (the dimension) and output a scale factor s and a rotation angle alpha (D-1 angles) that would fit exactly N samples of a Cartesian lattice in the unit box [0,1]^D.

Prerequisites: Math

Contact: Torsten Möller

Visualization for Optimization Problems

Can visualization beat traditional (offline) optimization problems? The goal of this project is to see how well visually guided optimization can compete with traditional optimization algorithms. Students will develop a visualization system to find optimum configurations of black box (i.e. unknown) algorithms from a contest.

Prerequisites: VIS, Mathematical Modeling
Programming languages: Javascript, (R or Python), and C++

Contact: Torsten Möller, Thomas Torsney-Weir

Visually Exploring Neural Networks

We have a collection of 100,000 different neural networks from the Tensorflow Playground. The core goal of this project is to create a visual interface to understand some of the basic properties of neural networks. Enabling a user to explore should help answer questions like the relationship of number of neurons and number of hidden layers, the impact of batch size, activation functions and other parameters on the quality of the network. Your tasks include:

  • fast prototyping with Tableau
  • getting familiar with the data set
  • querying neural network users on what parameters they want to explore (requirement analysis)
  • development of low-fi and high-fi prototypes

 

Prerequisites: VIS, FDA

Contact: Torsten Möller

Visually Analyzing the Fault Tolerance of Deep Neural Networks

The main objective is to design and implement a good and efficient way of visually investigating the resilience of deep neural networks against silent data corruption (bit flips) based on given empirical measurements. There are many possible causes for such faults (e.g., cosmic radiation, increasing density in chips, lower voltage which implies lower signal charge, etc.), and their "incidence" is expected to increase with current trends in chip architecture.

Starting point for the project is a given example data set which contains information about the relationship between single bit flips across various locations of a certain neural network (which layer, which neuron, which weight, which position within the floating-point representation of a real number, etc.) and the resulting accuracy of the network.

The task is to develop a tool which supports answering various questions about the influence of a bit flip on the resulting accuracy.

Examples for interesting questions are the following:

  • (empirical) distribution of the influence of a bit flip on the resulting accuracy over the positions in the floating-point representation
  • (empirical) distribution of the influence of a bit flip on the resulting accuracy over the layers in the network architecture
  • (empirical) distribution of the influence of a bit flip on the resulting accuracy over the weights in a given layer in the network architecture

In order to answer these questions, an iterative design process is required to

  • start with a requirement analysis (task & data analysis)
  • low-fi prototypes
  • high-fi prototypes
  • refinement
  • constant evaluation of the visual analysis tool.

The data set, the problem setting and the details of the requirements are provided by Prof. Gansterer, the supervision in visual analysis aspects is provided by Prof. Möller.

Prerequisites: VIS, FDA

Contact: Wilfried Gansterer, Torsten Möller

Data Visualization/Human Computer Interaction:

Analyzing Visual Literacy Test Performance and Exploring Alternatives

Visual literacy, the ability to interpret visual information, can be measured with dedicated tests. Students have completed such a test in the VIS course over the past 6-8 semesters. This project aims to analyze the collected results, examine factors influencing performance, and explore alternative approaches to assessing visual literacy. The analysis is envisioned to be carried out in Tableau.

  • Analyze students' test results from previous semesters
  • Compare performance across study backgrounds and semesters
  •  Compare students' performance before and after they took the VIS course
  • Investigate whether improvement correlates with other factors, such as overall grades
  • *Optional: Develop an alternative test fromat that could be integrated into the course in the future

Prerequisites: VIS

Contact: Antonia Saske, Torsten Möller, Laura Koesten

Data Analysis of Climate Visualization Perceptions

How do people perceive and interpret climate change visualizations? This project explores how individuals perceive and interpret data visualizations on climate change, based on data collected through an existing online survey. In this survey, participants provided insights into their experiences with visualizations, including reading accuracy, understandability, aesthetics, suggested improvements, or perceived take-away messages. You will develop your own research questions and explore the dataset to uncover patterns and relationships.

Key challenges:

  • Conducting exploratory data analysis of an existing survey dataset.

  • Identifying themes and key questions from survey data.

  • Performing quantitative analysis, including descriptive and inferential statistics.

  • Analyzing free-text responses (thematic analysis).

Prerequisites: FDA, VIS

Programming languages: Python or R

Contact: Regina Schuster, Laura Koesten

iTuner: Touch Interfaces for High-D Visualization

In order to understand simulations, machine learning algorithms, and geometric objects we need to interact with them. This is difficult to perform with something like a mouse which only has 2 axes of movement. Multitouch interfaces let us develop novel interactions for multi-dimensional data. The goals of this project are:

  • Develop a touch-screen interface for navigating high-dimensional spaces.
  • User interface designed for a tablet (ipad) to be used in concert with a larger screen such as a monitor or television.

Prerequisites: VIS, HCI

Contact: Torsten Möller, Thomas Torsney-Weir

LLMs as Explainers of Public AI Systems

Overview

This project explores whether LLMs can help novices to understand AI systems used in public institutions. It builds on XAI and HCI research into the information needs of different audiences and the adaptation of explanations to these needs. The focus is on both technical implementation and empirical validation (qualitative and/or quantitative) in user studies or interviews. A research paper could in principle be prepared and submitted.

Background

Public institutions increasingly use AI systems in decision-making, such as welfare fraud detection, student grading, and employment prediction. Understanding these systems is crucial for users, regulators, developers, and decision subjects. However, many systems lack transparency, making it difficult to trace and verify their decisions. Stakeholders should be able to understand how these systems are used (explanation) and for which purpose (justification). LLMs can flexibly process and present information but suffer from unreliability and lack of control. The project aims to study how useful an LLM can be as an adaptive explainer for different stakeholder audiences. For more information on the research background, you can consult the following papers: theory of XAI, XAI and stakeholder models, current challenges in XAI, public AI systems.

Aims and Expected Outcomes

The project aims to develop a high-fidelity prototype or functional web interface of an LLM to explain an AI system to various audiences. This involves requirement analysis, prototyping, validation through user studies, and a write-up (report, thesis, etc.). The process is flexible and guided by prior research, making it ideal for students interested in current research topics.

Suitable Candidates

Candidates should have an interest in AI explainability and user studies using HCI methods. Knowledge in web or interface development is advisable for the implementation. An interdisciplinary background in humanities or social sciences is also welcome.

Contact: Timothée Schmude, Torsten Möller

 

Objective Versus Subjective Measures of Interactions With Data Visualization

This project investigates the differences between objective and subjective evaluations of data visualizations, focusing on understanding and perceived understanding. You will conduct a study in which participants view a series of data visualizations and complete different tasks, including ones that measure their understanding. Participants will also provide feedback on perceived complexity, alongside other dimensions that describe their data visualization perception. You will also collect demographic information (age, sex) and experience factors (familiarity with data visualization, political leanings).

Key challenges include:

  • Comparing objective vs. subjective evaluations of complexity and understanding.

  • Designing an experimental set-up and conducting a study.

  • Analyzing how demographics and experience influence evaluations.

Prerequisites: VIS, FDA

Programming languages: Python or R

Contact: Laura Koesten, Torsten Möller

Visualizing Results From a Forced Choice Experiment

In forced-choice experiments, a participant is presented with several alternatives in a study in which stimuli are presented. The participant is forced to choose one stimulus over another (or over multiple) and cannot provide a neutral/custom response. 
This project will investigate how to visualize results from such experiments by exploring existing options and developing new or adapted visualization prototypes, which will be evaluated in a design study with users.

Prerequisites: Vis
Contact: Torsten Möller, Laura Koesten

HDI - Human Data Interaction

Explaining Descriptive Statistics

With Anscombe's Quartet [1] it was demonstrated quite figuratively that summary statistics can be very misleading or, at least, hard to interpret. Just recently, this example has become quite playful with the Dinozaur Dozen [2]. However, there are a number of statistical measures, that don't have an easy (visual) explanation. One of them is Krippendorf's alpha [3], a very common measure in the social science for measuring the agreement between subjective coders (as in labeling text or documents). The challenge of this project will be to:

  • understand the measure
  • develop simple alternatives
  • develop different visual representations that "bring this measure to life", i.e. make it easy(er) to understand

Prerequisites: VIS

[1] Anscombe, F. J. (1973). "Graphs in Statistical Analysis". American Statistician. 27 (1): 17–21. doi:10.1080/00031305.1973.10478966. JSTOR 2682899, see also Anscombe's quartet.

[2] Justin Matejka, George Fitzmaurice (2017), "Same Stats, Different Graphs: Generating Datasets with Varied Appearance and Identical Statistics through Simulated Annealing," ACM SIGCHI Conference on Human Factors in Computing Systems. see also Same Stats, Different Graphs. 

[3] Krippendorff, Klaus (1970). Estimating the reliability, systematic error, and random error of interval data. Educational and Psychological Measurement, 30 (1), 61–70. see also Krippendorff's alpha.

Explore Novel Interaction Techniques - Chatbots for Data Analysis

Chatbots are becoming more prevalent and are actively used by many companies. They offer a voice or text interface to interact with a computer. An example of a chatbot is amazon’s alexa, which can tell the time when asked.
The goal of this project is to find new possible ways to interact with an exploratory data analysis tool. Developing new interaction techniques would allow the user to explore and understand the data in a new fashion. For example, it could be possible to have a chat window next to a scatterplot that enables the user to enter queries such as: ‘show me the average’, which would then be reflected in the scatterplot.

  • Learn about natural language processing
  • Understand and compare interaction techniques
  • Develop a ‘conversation’ with a data analysis tool

Prerequisites: VIS, HCI

Contact: Torsten Möller

Multi-Level Emotional Measurement

This project focuses on improving the Self-Assessment Manikin (SAM) by integrating nuanced methods to capture emotional dimensions such as dominance, arousal, and valence. We aim to design a new representation of an emotion scale using:

  • Glyph-based representations: Exploring visual ways to represent complex emotional combinations.
  • Haptic feedback: Incorporating force-feedback to enable tactile interaction and support emotional expression.
  • Potentially: Physiological signals / Biometrics like heart rate or skin conductance 

You will explore decision trees as a method to guide users through emotional assessment, emphasizing intuitive, explainable AI-supported systems.  The goal is to measure subjective reactions and use these as inputs to adapt system interactions.

Key challenges include: 

  • Designing representations for complex emotional blends; addressing difficulties in defining and articulating emotions; 
  • Developing multimodal metrics to enhance user interaction and system adaptability. 

Background literature:

Margaret M Bradley and Peter J Lang. Measuring emotion: the self-assessment manikin and the semantic differential. Journal of behavior therapy and experimental psychiatry, 25(1):49–59, 1994.

Contact: Laura Koesten

Data Physicalization

3D Printing Alternatives for Data Physicalization

Not everyone has access to 3D printing, but physical data representations have existed since the beginning of mankind itself! Can we find fabrication methods that are feasible and accessible to a broad audience? What data are they suited for?

Contact: Daniel Pahr

Automation in Data Physicalization

Automation is a chance for physicalization to come to life! Arduino, Raspberry Pi and many more systems offer cheap and feasible ways to create small interactive devices. What kinds of sensors and motors can find applications to engage people with data representations?

Contact: Daniel Pahr

Dimensionality Reduction for Elastic 4D Printing

Elastic materials allow us to control the behavior of normally solid 3D printed objects. Volume data, i.e. scalar fields, as can be transformed into more or less faithful representations of the original object with creative fabrication pipelines. For multidimensional data with four or more dimensions, this method could help imitate dynamic behavior of objects, or generate tangible insights on abstract concepts such as flow and tensors.
 

Contact: Daniel Pahr

Sustainable Data Physicalization

Digital fabrication methods allow us to produce physical objects on demand, but with this process, additional challenges ensue. If we create complex objects they need to be aligned, and overhangs need to be dealt with in creative ways. How can we use a medium like 3D printing to create interactive, reusable data representations in a sustainable way?

Contact: Daniel Pahr

Sportsyanalytics

Sportsanalytics for Rowing Data

Within the AIROW (artificial intelligence in rowing) project (https://airow.univie.ac.at/) we are collaborating with sports scientists, the Austrian Rowing Federation, and other Data Scientists to collect, curate, and analyse data from top rowers on the training process, recovery, wellbeing and health.

Some of the main questions of the project is to better understand optimal training regimes as well as the impact of training, well-being, and other factors on the athlete's performance in order to optimize training load management.

Within this project, we have a number of student projects that aim at: 

  • Clustering and comparison of training sessions: Develop and evaluate methods for clustering and comparing training sessions to gain insights into athletic performance development.
  • Data analysis of rowing stroke force curves: Conduct an exploratory data analysis of force curves from rowing strokes to develop meaningful metrics and visualizations for athletes and coaches.
  • HCI for menstruation tracking: Design and implement a user interface to enable the input and visualization of crucial menstrual cycle information within the AIROW system, eliminating the need for a separate tracking application.
  • HCI for race data visualization: Develop a user-friendly interface focused on usability for importing and inputting test and race data, providing meaningful visualization and comparison options.
  • Menstruation and performance/well-being relations: Examine the connection between the menstrual cycle and an athlete's performance and well-being data.
  • Optimizing machine learning for performance prediction: Optimize existing machine learning algorithms by integrating additional information to enhance the accuracy of training performance predictions.
  • Relationship between external load and heart rate: Analyze the correlation between external loads, measured in watts, and heart rate characteristics during low-intensity endurance training sessions.
  • Understanding the link between external and internal load: Investigate the relationship between objective external training loads, such as power output or Training Impulse (TRIMP), and subjective internal physiological responses, measured through athletes' Rate of Perceived Exertion (RPE) scores.

In order to participate, you need to have passed (and enjoyed) the FDA course (Foundations of Data Analysis). The supervision will be done by various members of the project team. For more information, please contact Torsten M.

Prerequisites: FDA

Contact: Torsten Möller, Christoph Thiem, Judith Staudner