New loss functions enhance ML model robustness to outliers

Researchers have developed new loss functions and regression models designed to improve robustness against outliers in machine learning datasets. The proposed approach modulates the learning rate based on outlier sensitivity, with a focus on alternate loss functions that more closely approximate absolute error than existing methods like Huber and log-cosh losses. The paper introduces Square Root Loss (SRL) and Smooth Mean Absolute Error (SMAE) as superior alternatives, demonstrating their effectiveness through comparisons on various benchmarks and presenting new robust linear regression models with vectorized parameter update formulas optimized for GPUs. AI

IMPACT Improves the reliability of machine learning models in real-world scenarios with noisy data.

RANK_REASON Academic paper detailing new methods and models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• graphics processing unit
• Huber loss function
• log-cosh loss
• Mathew M Noel
• Smooth Mean Absolute Error
• Square Root Loss

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →