Adapting Multi-objectivized Software Configuration Tuning to Overcome Local Optima and Improve Efficiency

The authors propose an adaptive multi-objectivization (AdMMO) method to overcome the limitations of the fixed weight in the original multi-objectivization (MMO) approach for software configuration tuning. AdMMO dynamically adjusts the weight during the tuning to maintain an appropriate proportion of unique nondominated configurations, which serves as a better indicator of the ideal balance between exploitation and exploration. Additionally, AdMMO employs a partial duplicate retention mechanism to handle the issue of too many duplicate configurations without losing the rich information provided by the "good" duplicates.