Sparse Logistic Regression With L _1/2 Penalty for Emotion Recognition in Electroencephalography Classification

Area of Science:

• Neuroscience
• Machine Learning
• Signal Processing

Background:

• Emotion recognition using electroencephalography (EEG) is crucial for brain-computer interfaces.
• Classifying EEG data is challenging due to noise and large datasets.
• Effective feature extraction is vital for accurate EEG signal processing.

Purpose of the Study:

• To investigate the efficacy of L1/2 penalty in sparse logistic regression for three-classification EEG emotion recognition.
• To compare L1/2 penalty with existing regularization methods like L1, Ridge Regression, and Elastic Net.
• To demonstrate the benefits of L1/2 regularization for high-dimensional, small-sample EEG data.

Main Methods:

• Implemented L1/2 penalty logistic regression using a coordinate descent algorithm.
• Employed a univariate semi-threshold operator for L1/2 penalty logistic regression.
• Evaluated the proposed method on both simulated and real EEG data.

Main Results:

• The proposed L1/2 penalty logistic regression achieved higher classification accuracy compared to L1, Ridge Regression, and Elastic Net.
• The method effectively extracts fewer, more informative EEG signals.
• Demonstrated improved computational accuracy and reduced complexity.

Conclusions:

• Sparse logistic regression with L1/2 penalty is an effective technique for EEG emotion recognition.
• This method offers significant advantages for high-dimensional and small-sample EEG datasets.
• The findings support the use of L1/2 penalty for practical emotion recognition applications.