Navigation auf uzh.ch
Level: MA
Responsible Person: Dr. Taehoon Kim
Keywords: Graph representation learning, Machine learning
This thesis project titled "Basic Graph Reinforcement Learning using Blockchain Data" aims to explore the application of graph-based reinforcement learning techniques to analyse and derive insights from blockchain network structures. This project would utilise the inherent graph-like nature of blockchain data, where transactions and addresses form nodes and edges in a complex network.The student would begin by collecting and preprocessing blockchain data, likely focusing on a specific blockchain like Ethereum or Bitcoin. They would then construct a graph representation of the blockchain, with addresses as nodes and transactions as edges. The reinforcement learning component could involve training agents to navigate this graph structure, potentially to identify patterns, anomalies, or optimize certain network behaviors.Key aspects of the project might include:
Implementing graph construction algorithms tailored to blockchain data
Designing appropriate state and action spaces for the reinforcement learning agents (à The most important point of this project)
Developing reward functions that capture meaningful blockchain network properties
Experimenting with various graph reinforcement learning algorithms, such as Graph Neural Networks (GNNs) combined with Q-learning or policy gradient methods
Evaluating the performance of the learned models in tasks such as transaction volume prediction, address clustering, or detecting unusual network activities
This project would contribute to the growing field of blockchain analytics while also advancing the application of reinforcement learning to graph-structured data. The insights gained could have implications for blockchain security, network optimization, and economic analysis of cryptocurrency ecosystems.
References:
Wu, Y., et al. (2021). A Survey of Graph Neural Networks for Recommender Systems: Challenges, Methods, and Directions. arXiv:2109.12843. Retrieved from https://arxiv.org/abs/2109.12843
Xie, Y., et al. (2021). A Survey on Graph Neural Networks and Graph Transformers in Computer Vision: A Task-Oriented Perspective. arXiv:2109.00901. Retrieved from https://arxiv.org/abs/2109.00901
Jiang, J., et al. (2020). Graph Convolutional Reinforcement Learning for Multi-Agent Cooperation. arXiv:1810.09202. Retrieved from https://arxiv.org/abs/1810.09202
Chen, Z., et al. (2020). GCC: Graph Contrastive Coding for Graph Neural Network Pre-Training. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (pp. 1150-1160).
Zheng, L., et al. (2020). MaGNN: Metapath Aggregated Graph Neural Network for Heterogeneous Graph Embedding. In Proceedings of The Web Conference 2020 (pp. 2331-2341).
Zhang, Y., et al. (2024). Machine Learning for Blockchain Data Analysis: Progress and Opportunities. arXiv:2404.18251v1. Retrieved from https://arxiv.org/html/2404.18251v1
Albshaier, L., Almarri, S., & Hafizur Rahman, M.M. (2024). A Review of Blockchain's Role in E-Commerce Transactions: Open Challenges, and Future Research Directions. Computers, 13(1), 27. https://doi.org/10.3390/computers13010027