Using Graph Neural Networks To Embrace The Dependency In Your Data by Usman Zafar - GoDataFest 2022
0 likes117 views
The document discusses the application of graph neural networks (GNNs) in embracing data dependencies, highlighting their capabilities in problem-solving. It presents a use case of predicting demand in new train stations and covers various graph-related concepts, including node and link prediction. Additionally, it mentions open-source tools like PyTorch Geometric for implementing GNNs in practice.
![Performance
43
[1] Cui et al, Traffic Graph Convolutional Recurrent Neural Network: A Deep Learning Framework for Network-Scale Traffic Learning and Forecasting, 2018](https://image.slidesharecdn.com/presentation-221028132733-b99772d4/85/Using-Graph-Neural-Networks-To-Embrace-The-Dependency-In-Your-Data-by-Usman-Zafar-GoDataFest-2022-43-320.jpg)
Using Graph Neural Networks To Embrace The Dependency In Your Data by Usman Zafar - GoDataFest 2022
- 1. Using Graph Neural Networks To Embrace The Dependency In Your Data Usman Zafar GoDataDriven
- 2. Introduction l Machine Learning Engineer @ GoDataDriven l Background in data science l Fascinated by graphs and their problem solving abilities 2
- 3. Agenda 3 l Example use-case l Common methods and their limitations l How graphs can save the day l Open-source tools to get started
- 6. Example use-case 6 l Predicting/categorising demand in new train stations
- 13. Framing the problem...using a graph 13
- 14. Encoding information in the edges 14
- 15. Encoding information in the nodes 15
- 16. Graph Labels 16
- 22. What Now? 22
- 25. Laplacian 25 L = D - A
- 34. The Mathematics 34 Convolving a feature vector
- 43. Performance 43 [1] Cui et al, Traffic Graph Convolutional Recurrent Neural Network: A Deep Learning Framework for Network-Scale Traffic Learning and Forecasting, 2018
- 44. Graph Database 44 Traditional SQL Query Graph Query
- 45. Spark GraphX 45
- 47. Q&A 47