Estimating the Event-Related Potential from Few EEG Trials
Authors: Anders Vestergaard Nørskov, Kasper Jørgensen, Alexander Neergaard Zahid, Morten Mørup
TMLR 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate our approach in the challenging zero-shot scenario of generalizing to new subjects considering three different publicly available data sources; i) the comprehensive ERP CORE dataset that includes over 50,000 EEG trials across six ERP paradigms from 40 subjects, ii) the large P300 Speller BCI dataset, and iii) a neuroimaging dataset on face perception consisting of both EEG and MEG data. We consistently find that our method in the few trial regime provides substantially better ERP estimates than commonly used conventional and robust averaging procedures. |
| Researcher Affiliation | Collaboration | Anders Vestergaard Nørskov EMAIL Department of Applied Mathematics and Computer Science, Technical University of Denmark... Alexander Neergaard Zahid EMAIL Department of Applied Mathematics and Computer Science, Technical University of Denmark WS Audiology, Lynge, Denmark |
| Pseudocode | No | The paper describes the methodology in text and mathematical equations but does not present any clearly labeled pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code is available at https://github.com/andersxa/EEG2ERP. |
| Open Datasets | Yes | We developed and tested EEG2ERP on the ERP CORE dataset (Kappenman et al., 2021)... Additionally, we investigated EEG2ERP on an EEG and MEG dataset originating from face perception experiments (Wakeman & Henson, 2015). Finally, we evaluated EEG2ERP on the P300 BCI Speller dataset (Won et al., 2022). |
| Dataset Splits | Yes | To this end, the ERP CORE dataset was split such that trials from 28 subjects were used for training, 4 for validation, and the remaining 8 for testing. The Wakeman-Henson EEG and MEG datasets were divided into 11 subjects for training, 2 for validation, and 3 for testing. Finally, the P300 BCI Speller dataset was split into 38 subjects for training, 5 for validation and 12 for testing. The exact subject splits are provided in Appendix I.1. |
| Hardware Specification | No | The paper does not provide specific hardware details such as GPU/CPU models, processors, or memory used for running its experiments. |
| Software Dependencies | No | The paper mentions 'All M/EEG recordings were preprocessed in Field Trip (Oostenveld et al., 2011).' and 'We used the XGBoost Classifier (Chen & Guestrin, 2016)' but does not provide specific version numbers for these or any other ancillary software components used in their implementation. |
| Experiment Setup | Yes | We train EEGERP for Emax = 200 epochs, annealing the predicted standard deviation over the first Etarget = 100 epochs, and employ a learning rate of 4 10 4 with a One Cycle learning-rate scheduler. Complete parameter values and experimental settings are detailed in the associated source code, which is provided as supplementary material and hosted at https://github.com/andersxa/EEG2ERP. |