Multiscale Deep Learning Convolutional Neural Network for ADHD Detection Using EEG: Source Code and Ablation Logs

This record provides the cleaned EEG-MSCNet source code and ablation experiment logs supporting the manuscript “Multiscale Deep Learning Convolutional Neural Network for ADHD Detection Using EEG.” EEG-MSCNet is a multiscale convolutional neural network for pediatric electroencephalography (EEG)-based Attention-Deficit/Hyperactivity Disorder (ADHD) classification under subject-level leave-one-subject-out (LOSO) evaluation. The archive includes the experiment launcher, model definitions, configuration files, analysis utilities, tests, Docker execution files, and ablation logs for MultiScaleCNN_ID_1 through MultiScaleCNN_ID_55. These logs support the manuscript analyses of preprocessing, architectural variants, channel-region configurations, computational complexity, stability across repeated runs, and additional exploratory window-overlap experiments. The original EEG dataset is not redistributed in this archive. Users must obtain the dataset from its original public source and follow the corresponding access and usage terms. This archive is intended to support reproducibility, transparent comparison, and reuse of the EEG-MSCNet experimental pipeline.