Deep Transformer Network and CNN Model with About 200k Parameters to Classify P300 EEG Signal

Abstract

This research gathers participants“ electroencephalographic (EEG) data to build classifiers that can decode users” mental states. Although deep learning models can incorporate domain-dependent feature extraction into the classifier design, the architecture selection process for BCIs is often based on domain expertise. This paper examines whether it can build reliable classifiers for interpreting P300 event-related potentials using a systematic model selection mixed with common deep learning architectures. We report the findings of a CNNTransformer network model with 50k and 200k parameters in particular. We systematically investigate if a hybrid architecture which is Transformer and CNN, can improve the classification performance of P300 datasets. Because we would like to know how to improve the model selection process for P300 signals data analysis. To help the readers understand that the best model selection strategy with more muscular models than the existing ones may make the lives of stroke patients more convenient. If our readers are targeting to know the modern muscular deep learning architectures to analyze their recorded P300 dataset, then we recommend reading this paper.

More by Zhenis Otarbay

• Ocular-correlated signals and decision rules in DTW decoding of motor-imagery EEG: code and resultsZenodo · 2026
• When Deployment Changes a Learned Action: Post-Inference Projection in Offline Multi-Agent Reinforcement LearningAI · 2026
• Robust Offline Multi-Agent Reinforcement Learning for Latency-Aware SDN Path Control in 6G-Oriented Network SoftwarizationTechnologies · 2026 · Citations: 1
• An Exploratory Study of Gender Effects on Pupillary Cognitive Load in Human–Robot InteractionIEEE International Conference on Smart Information Systems and Technologies · 2026