Optimization of Deep Architectures for EEG Signal Classification: An AutoML Approach Using Evolutionary Algorithms

Area of Science:

• Neuroscience
• Machine Learning
• Signal Processing

Background:

• Electroencephalography (EEG) signal classification is complex due to low signal-to-noise ratios and artifacts.
• Traditional methods rely on predefined features, limiting performance across diverse experimental conditions.
• Deep learning offers potential for improved EEG classification without manual feature engineering.

Purpose of the Study:

• To develop a method for optimizing deep learning models for EEG signal classification.
• To enhance classification performance by optimizing both hyperparameters and network architecture.
• To improve computational and energy efficiency of deep learning models for EEG analysis.

Main Methods:

• A novel optimization method for deep learning models was proposed.
• The method optimizes deep learning architecture, including layer combinations and hyperparameters.
• Experimental validation was performed using EEG datasets.

Main Results:

• Optimized deep architectures significantly outperformed baseline approaches in EEG classification.
• The proposed method resulted in computationally efficient models.
• Optimized architectures demonstrated improved energy efficiency compared to baseline models.

Conclusions:

• The developed optimization method effectively enhances deep learning model performance for EEG classification.
• The approach offers a pathway to more accurate, efficient, and energy-saving EEG analysis.
• This work contributes to advancing machine learning applications in neuroscience and biomedical signal processing.