RMoR-Aion: Robust Multioutput Regression by Simultaneously Alleviating Input and Output Noises.

Area of Science:

• Machine Learning
• Data Science
• Statistical Modeling

Background:

• Multioutput regression models predict multiple continuous variables simultaneously, capturing inter-variable correlations.
• Low-rank methods are mainstream for multioutput regression, reducing parameters and enhancing performance.
• Existing low-rank methods are vulnerable to noise in input and output data.

Purpose of the Study:

• To develop a novel robust multioutput regression method (RMoR-Aion) that simultaneously alleviates input and output noises.
• To enhance the stability and accuracy of multioutput regression in noisy environments.
• To leverage auxiliary matrices and output manifold constraints for noise reduction.

Main Methods:

• Developed RMoR-Aion, a novel multioutput regression technique.
• Utilized auxiliary matrices to exploit and alleviate both input and output noises.
• Incorporated a prediction output manifold constraint, using output variable correlations to mitigate noise effects.

Main Results:

• RMoR-Aion demonstrated superior effectiveness compared to state-of-the-art baseline methods in empirical studies.
• The proposed method showed enhanced stability, particularly in scenarios with artificial noise.
• Simultaneous alleviation of input and output noises proved beneficial for multioutput regression performance.

Conclusions:

• RMoR-Aion offers a robust solution for multioutput regression problems affected by noise.
• The method provides improved prediction accuracy and stability over existing techniques.
• Leveraging noise information and manifold constraints is a promising direction for robust multioutput regression.