This Master’s thesis investigates the application of Symbolic Regression (SR) to electroencephalogram (EEG) time series in order to characterize the complex, non-linear dynamics of the brain. While modern Deep Learning models possess high predictive power for EEG analysis, their ”black-box” nature fails to provide the strict interpretability required to uncover the underlying neurophysiological mechanisms driving the brain. Symbolic Regression attempts to bridge this gap by autonomously extracting explicit, interpretable mathematical equations directly from data. Using the PySR algorithm and a 64-electrode EEG dataset from healthy control subjects, this study evaluates the capability of SR to map both intra-channel neural dynamics and inter-channel spatial connectivity between electrodes.
Tribunal:
Miguel Cornelles Soriano (president)
Leonardo Gollo Lyra (secretary)
Massimiliano Zanin (vocal)
Detalles de contacto:
Massimiliano Zanin Contact form