Grape Up has researched Federated Learning in the automotive industry, examining its advantages, performance, requirements, and risks.
Exploring Federated Learning for automotive software development
Federated Learning is a machine learning concept based on using decentralized data to build a global AI model.
The advantage of using decentralized data, compared to a centralized system, is that the data never leaves the edge device where it was collected. This makes it very convenient for use cases requiring sensitive customer data to be collected and used for the training process.
As an automotive software development company, Grape Up has focused its research on Federated Learning in the following core aspects:
- Ability to use it in the automotive industry – advantages over currently used solutions, performance implications, requirements, and potential risks.
- Technical research by our internal Data Science team on the existing tools and frameworks for their usability and features in terms of splitting the work and combining results.
- Research of related work, mainly in scientific papers, focused primarily on the research being done for automotive or similar automotive use cases, and their results, especially in terms of performance, scalability, and accuracy.