Correntropy-Based Logistic Regression With Automatic Relevance Determination for Robust Sparse Brain Activity...

Area of Science:

• Neuroscience
• Machine Learning
• Signal Processing

Background:

• Sparse classification models are used for brain activity decoding but suffer from noise in recordings.
• Existing methods often experience performance degradation due to inherent noise in electroencephalogram (EEG) and functional magnetic resonance imaging (fMRI) data.

Purpose of the Study:

• To propose a novel robust and sparse classification algorithm to address noise-induced performance degradation.
• To enhance the accuracy and reliability of brain activity decoding for mental state and intention prediction.

Main Methods:

• Introduced the correntropy learning framework into an automatic relevance determination-based sparse classification model.
• Developed a new correntropy-based robust sparse logistic regression algorithm.
• Evaluated the algorithm on synthetic, EEG, and fMRI datasets.

Main Results:

• The proposed algorithm achieved higher classification accuracy on noisy, high-dimensional datasets.
• The method effectively selected more informative features for brain activity decoding tasks.
• Experimental results confirmed superior performance compared to existing methods.

Conclusions:

• Integrating correntropy learning with automatic relevance determination significantly improves robustness against noise.
• The developed algorithm offers a more powerful and adequate approach for robust sparse brain decoding.
• This advancement has significant implications for real-world brain activity decoding and brain-computer interfaces.