Contact: dkelesis@iit.demokritos.gr
Graph-based auction (or in general auctions using GNNs) mechanisms use the structure of graphs to model complex bidding environments, such as those found in networked markets or combinatorial auctions. Current approaches leverage Vickrey-Clarke-Groves (VCG) mechanisms and greedy algorithms for combinatorial auctions. However, these methods often struggle with scalability and efficiency in large, complex networks.
This thesis will explore the development of scalable graph neural network (GNN)-based auction algorithms. Research questions include: How can GNNs be utilized to model and solve large-scale auction problems? What are the performance gains in terms of computational efficiency and solution quality? Can GNN-based methods provide better approximations for optimal auction outcomes in large networks? Try to use GNNs to predict what bidders will bid and/or what price will they diffuse.
Relevant literature:
[1] https://arxiv.org/pdf/2009.13697
[2] https://www.ijcai.org/proceedings/2019/0062.pdf