| dc.description.abstract | This paper explores the application of machine learning techniques to classify cognitive processing operations using EEG data, contributing to the fields of cognitive neuroscience and machine
learning (ML). The research discussed is rooted in the theory of processing operations, examining
how the brain manages tasks in sequence.
Our methods involve the use of different machine learning algorithms, such as convolutional
neural networks (CNNs), recurrent neural networks (RNNs) and transformer networks, to analyze
complex patterns in electroencephalography (EEG) data. Our methods are supported by a flexible
framework for data collection, processing, and model training, enabling adaptability and integration of new insights. The goal of this research is to develop a proof-of-concept for finding similar
cognitive processing operations across different contexts, using the supposed onsets of these operations as provided by hidden multivariate pattern (HMP) analysis.
We showcase parameter tests aimed at identifying effective strategies for applying machine
learning to EEG data analysis. The results demonstrate the utility of machine learning in decoding
brain functioning and answer some questions around which techniques to use in analyzing EEG
data within the context of this paper.
The results show that all three previously mentioned classes of machine learning models are
able to generalize very well across condition (± 1% loss in performance), slightly worse across lab
(± 30% loss in performance), but do not generalize across task. We believe that these results show
that generalizing across contexts is principally possible and can function as a first step into discerning what makes each cognitive processing operation unique. | |
