How to "DODGE" Complex Software Analytics?
This addresses efficiency bottlenecks in applying machine learning to software engineering tasks, though it appears incremental as it optimizes an existing process rather than introducing a new paradigm.
The paper tackles the problem of slow hyperparameter optimization in software analytics by identifying and ignoring redundant tunings that produce indistinguishable results. Their DODGE tool runs orders of magnitude faster while generating more accurate predictions than prior state-of-the-art approaches.
Machine learning techniques applied to software engineering tasks can be improved by hyperparameter optimization, i.e., automatic tools that find good settings for a learner's control parameters. We show that such hyperparameter optimization can be unnecessarily slow, particularly when the optimizers waste time exploring "redundant tunings"', i.e., pairs of tunings which lead to indistinguishable results. By ignoring redundant tunings, DODGE, a tuning tool, runs orders of magnitude faster, while also generating learners with more accurate predictions than seen in prior state-of-the-art approaches.