Understanding The Effectiveness of Lossy Compression in Machine Learning Training Sets
This addresses the challenge of handling large training data volumes in HPC and edge computing for ML/AI practitioners, providing systematic guidance for compression use and design.
The paper tackles the problem of how lossy compression affects machine learning model quality by conducting a comprehensive evaluation of 17 data reduction methods on 7 ML/AI applications, showing that modern lossy compression can achieve 50-100x compression ratios with less than 1% loss in quality.
Learning and Artificial Intelligence (ML/AI) techniques have become increasingly prevalent in high performance computing (HPC). However, these methods depend on vast volumes of floating point data for training and validation which need methods to share the data on a wide area network (WAN) or to transfer it from edge devices to data centers. Data compression can be a solution to these problems, but an in-depth understanding of how lossy compression affects model quality is needed. Prior work largely considers a single application or compression method. We designed a systematic methodology for evaluating data reduction techniques for ML/AI, and we use it to perform a very comprehensive evaluation with 17 data reduction methods on 7 ML/AI applications to show modern lossy compression methods can achieve a 50-100x compression ratio improvement for a 1% or less loss in quality. We identify critical insights that guide the future use and design of lossy compressors for ML/AI.