Semi-proximal ADMM for fused Lasso penalized least absolute deviation in partially linear model

Key Points

• The semi-proximal ADMM algorithm significantly enhances prediction accuracy when applied to fused lasso outcomes.
• For the penalized least absolute deviation approach, the algorithm reduces biases in parameter estimates across diverse datasets.
• This work assesses a novel algorithm within the framework of partially linear models to address existing estimation weaknesses.
• Improvements in predictive modeling could have important implications, particularly in areas demanding precise parameter estimation.