1326. Federated deep preference learning for Multiple Criteria Decision Aiding
Invited abstract in session MA-8: Advances in preference learning methods, stream Multiple Criteria Decision Aiding.
Monday, 8:30-10:00Room: Clarendon SR 2.08
Authors (first author is the speaker)
| 1. | Krzysztof Martyn
|
| Institute of Computing Science, Poznan University of Technology | |
| 2. | Milosz Kadzinski
|
| Institute of Computing Science, Poznan University of Technology |
Abstract
In this study, we propose novel approaches for developing collaborative decision models that involve multiple decision-makers. Each of them possesses private alternatives and preference information that they either cannot or prefer not to share. Consequently, local models are trained using these private datasets and are never exchanged between decision-makers. Instead, only model-related information and learning parameters are transmitted during training to a central coordinator, who then constructs a comprehensive model subsequently evaluated by all decision-makers. The proposed framework has been validated using neural network-based UTADIS and outranking methods applied to the threshold-based sorting problem, which employs holistic preference information to assign alternatives to classes. We explore various strategies for developing a shared decision-making approach, balancing different levels of data security with the time required to reach a consensus. Extensive experiments on large datasets demonstrate that our approach effectively preserves privacy while facilitating collaboration, thereby offering a viable solution for real-world decision-making scenarios.
Keywords
- Decision Support Systems
- Group Decision Making and Negotiation
- Machine Learning
Status: accepted
Back to the list of papers