Graph data analysis using deep learning methods
Thesis title in Czech: | Analýza grafových dat pomocí metod hlubokého učení |
---|---|
Thesis title in English: | Graph data analysis using deep learning methods |
Key words: | grafové embeddingy, deep learning, vizualizace grafu, rekonstrukce grafu, predikce hran |
English key words: | graph embedding, deep learning, graph visualisation, graph reconstruction, link prediction |
Academic year of topic announcement: | 2018/2019 |
Thesis type: | diploma thesis |
Thesis language: | angličtina |
Department: | Department of Software Engineering (32-KSI) |
Supervisor: | RNDr. Martin Svoboda, Ph.D. |
Author: | Mgr. Vladislav Vancák - assigned and confirmed by the Study Dept. |
Date of registration: | 30.10.2018 |
Date of assignment: | 31.10.2018 |
Confirmed by Study dept. on: | 30.11.2018 |
Date and time of defence: | 10.06.2019 09:00 |
Date of electronic submission: | 08.05.2019 |
Date of submission of printed version: | 10.05.2019 |
Date of proceeded defence: | 10.06.2019 |
Opponents: | Mgr. Vladan Majerech, Dr. |
Guidelines |
Modern real-world applications often rely on data represented in graph structures rather than relational tables. For example, the financial sector is a rich source of transaction-based data (money transfers, card payments, cash withdrawals, etc.) which can be transformed into graph structures describing both direct and indirect relationships or similarities between individual clients. One of the most promising approaches for interpreting such data structures for the purpose of machine learning and data analysis are various graph embedding methods where we are attempting to find suitable representations of graph nodes in a vector space.
The goal of this thesis is to analyze and mutually compare recent deep learning graph embedding approaches, characterize their features and identify their advantages and disadvantages. Selected approaches will be implemented so that they can also be experimentally evaluated and compared in the context of banking, insurance or other well-chosen industry relevant predictive tasks, and thus their applicability to real-world problems can be examined and interpreted. |
References |
Palash Goyal, Emilio Ferrara - Graph Embedding Techniques, Applications, and Performance: A Survey - https://arxiv.org/abs/1705.02801
Aditya Grover, Jure Leskovec - node2vec: Scalable Feature Learning for Networks - https://arxiv.org/abs/1607.00653 Ke Tu, Peng Cui, Xiao Wang, Fei Wang, Wenwu Zhu - Structural Deep Embedding for Hyper-Networks - https://arxiv.org/abs/1711.10146 Ian Goodfellow, Yoshua Bengio, Aaron Courville - Deep Learning, MIT Press, 2016 - http://www.deeplearningbook.org |