We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Practical Geometric Deep Learning in Python • Pantelis Elinas • YOW! 2019
Learn how to analyze connected data with StellarGraph, a Python library for graph data processing, and explore GCN and Graphsage methods for node classification and link prediction.
- The speaker introduces StellarGraph, a Python library for graph data processing.
- The goal is to help users analyze connected data, focusing on relationships.
- Graphs are a powerful representation of connected data, making it possible to analyze complex relationships.
- GCN (Graph Convolutional Network) is introduced as a method to solve graph-related problems.
- Graphsage is another method that can be used, offering good performance and scalability.
- Node classification is a problem where you want to predict an attribute of a node based on its features.
- Traditional methods struggle with this problem, but GCN and Graphsage can be used for better results.
- The graph dataset being used is a Twitter dataset, where users are classified as either hateful or not.
- GCN defines a new type of neural network layer, applicable to graphs.
- Graphsage defines a new type of neural network layer that can be used for graph processing.
- Sampling is used to create a smaller graph for processing.
- The number of samples is the size of the neighborhood for each layer.
- Neural networks can be used for graph processing, and Keras is a good library to use.
- StellarGraph offers a consistent API and is highly scalable.
- Link prediction is also a problem where you try to predict the relationships between nodes in a graph.
- Manual feature engineering might not be the best way to solve these problems.
- Instead, StellarGraph can help with creating features automatically.
- The library is highly extensible, allowing users to add their own algorithms.
- The library is also highly scalable, making it suitable for large datasets.
- Data61 is behind StellarGraph, and the team is always looking for contributors.