Hands-On Graph Neural Networks using Python

Only ten years after their creation, graph neural networks have become one of the most interesting architectures in deep learning. They have revolutionized multi-billion-dollar industries like drug discovery, where they predicted a brand-new antibiotic named Halicin. Tech companies are now trying to apply them everywhere: recommender systems for food, videos, and romantic partners; fake news detection, chip design, and 3D reconstruction.

In Graph Neural Networks, we will explore the fundamentals of graph theory and create our own datasets from raw or tabular data. We will introduce major graph neural network architectures to understand crucial concepts like graph convolution and self-attention. This knowledge will then be applied to understand and implement more specialized models, designed for various tasks (including link prediction and graph classification) or contexts (spatio-temporal data, heterogeneous graphs, and so on). Finally, we will solve real-life problems using this technology and start building a professional portfolio.

By the end of this book, you will become a Graph Neural Network expert. You will be able to reframe your problems to leverage the unreasonable effectiveness of this architecture. With these skills, you will create unique solutions using novel, state-of-the-art approaches.

What you will learn

• Create your own graph datasets from tabular or raw data
• Transform nodes and edges into high-quality embeddings
• Implement graph neural networks using PyTorch Geometric
• Select the best graph neural network model according to your problem
• Perform tasks like node classification, graph generation, link prediction
• Apply this knowledge to real use cases with raw data

• # Computer Neural Networks