Building Stable Kernel Trees with Machine Learning

Explore the application of machine learning in building stable kernel trees in this Linux Plumbers Conference talk. Delve into the process of identifying and reviewing fixes, automating patch selection, and the challenges of using neural networks for this task. Learn about the different approaches to stable tree maintenance, including the use of stable tags and the importance of understanding what constitutes a fix. Discover how convolutional neural networks, typically used in image processing, are adapted for code structure analysis. Gain insights into the current results, future work possibilities, and the potential for involving non-developers in kernel maintenance. Engage with the speakers' explanations, examples, and their vision for improving the stable kernel development process through innovative machine learning techniques.

Introduction
How do stuff get in stable
What is a fix
Why Stable Trees
Stable Tags
Fix Differently
Reviewing Patches
Automating Patches
Unbalanced Talk
Neural Network
Problems with Neural Network
Not all fixes are stable
Commits to Stable
Conclusions
Explanation
Example
Wellknown Developers
Neural Networks
Training Data
How can we improve
Convolutional Neural Network
Image Processing
MaxPooling
Text
Natural Language
Representation
Dropping Stop Words
Code Structure
Results
Future work
Bug fixes
Non developers as maintainers
Propagation
Questions