Markus Utke

h-index3
2papers
50citations

2 Papers

3.6GTJun 22
Mixed Voting Rules for Participatory Budgeting

Anton Baychkov, Markus Brill, Markus Utke

Designing and analyzing voting rules for Participatory Budgeting (PB) elections is an active research area in computational social choice. Many PB voting rules aim to optimize a specific objective. For instance, the ubiquitous Greedy rule attempts to maximize utilitarian welfare, while the Method of Equal Shares (MES) aims to achieve proportional representation. However, it is often desirable to achieve good outcomes on multiple objectives rather than a close-to-perfect outcome for one. Inspired by mixed-member systems for parliamentary elections, we introduce mixed voting rules for PB. These are composed of a sequence of two or more rules that can each spend some fraction of the overall budget in order to add projects to the set selected by earlier rules. We develop a theoretical framework for formulating and analyzing mixed PB voting rules, and explore how existing rules can be adapted to this framework. We particularly focus on MES and its potential to address imbalances in representation created by earlier rules. We propose different ways to adjust MES voter budgets based on how satisfied voters are with previously chosen projects, and examine how well the resulting rules approximate well-known proportionality axioms such as EJR+. In particular, we show that one of these methods improves upon a natural proportionality baseline. We also extend our main positive result to general additive satisfaction functions. We complement our theoretical results with an extensive empirical analysis of real-world PB elections. Our experiments show that mixed rules can achieve favorable trade-offs between utilitarian welfare and proportionality. We identify several refinements that further improve their performance, and apply our framework to PB rules beyond Greedy and MES.

4.3MMMay 2, 2020
Towards Deep Learning Methods for Quality Assessment of Computer-Generated Imagery

Markus Utke, Saman Zadtootaghaj, Steven Schmidt et al.

Video gaming streaming services are growing rapidly due to new services such as passive video streaming, e.g. Twitch.tv, and cloud gaming, e.g. Nvidia Geforce Now. In contrast to traditional video content, gaming content has special characteristics such as extremely high motion for some games, special motion patterns, synthetic content and repetitive content, which makes the state-of-the-art video and image quality metrics perform weaker for this special computer generated content. In this paper, we outline our plan to build a deep learningbased quality metric for video gaming quality assessment. In addition, we present initial results by training the network based on VMAF values as a ground truth to give some insights on how to build a metric in future. The paper describes the method that is used to choose an appropriate Convolutional Neural Network architecture. Furthermore, we estimate the size of the required subjective quality dataset which achieves a sufficiently high performance. The results show that by taking around 5k images for training of the last six modules of Xception, we can obtain a relatively high performance metric to assess the quality of distorted video games.