KeystoneML: Optimizing Pipelines for Large-Scale Advanced Analytics
This system addresses the problem of high resource requirements and complexity in large-scale ML applications for data scientists and engineers, offering incremental improvements in performance and usability over existing systems.
The authors tackled the challenge of optimizing end-to-end machine learning pipelines for large-scale analytics by developing KeystoneML, a system that improves training throughput by up to 15x on real-world datasets like image classification.
Modern advanced analytics applications make use of machine learning techniques and contain multiple steps of domain-specific and general-purpose processing with high resource requirements. We present KeystoneML, a system that captures and optimizes the end-to-end large-scale machine learning applications for high-throughput training in a distributed environment with a high-level API. This approach offers increased ease of use and higher performance over existing systems for large scale learning. We demonstrate the effectiveness of KeystoneML in achieving high quality statistical accuracy and scalable training using real world datasets in several domains. By optimizing execution KeystoneML achieves up to 15x training throughput over unoptimized execution on a real image classification application.