Academic Journal
A Few-Layer Multilayer Perceptron is Worth Attention for EEG Classification in Rapid Serial Visual Presentation Task.
| Title: | A Few-Layer Multilayer Perceptron is Worth Attention for EEG Classification in Rapid Serial Visual Presentation Task. |
|---|---|
| Authors: | Zhang Z; School of Electronic Engineering, Xidian University, Xi'an 710071, Shaanxi, P. R. China., Zheng Y; School of Electronic Engineering, Xidian University, Xi'an 710071, Shaanxi, P. R. China., Guo K; School of Electronic Engineering, Xidian University, Xi'an 710071, Shaanxi, P. R. China., Liang J; School of Electronic Engineering, Xidian University, Xi'an 710071, Shaanxi, P. R. China., Dong M; School of Life Science and Technology, Xidian University, Xi'an 710126, Shaanxi, P. R. China. |
| Source: | International journal of neural systems [Int J Neural Syst] 2026 Sep; Vol. 36 (9), pp. 2650030. Date of Electronic Publication: 2026 Apr 10. |
| Publication Type: | Journal Article |
| Language: | English |
| Journal Info: | Publisher: World Scientific Pub. Co Country of Publication: Singapore NLM ID: 9100527 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1793-6462 (Electronic) Linking ISSN: 01290657 NLM ISO Abbreviation: Int J Neural Syst Subsets: MEDLINE |
| Imprint Name(s): | Original Publication: Singapore ; Teaneck, N.J. : World Scientific Pub. Co., c1989- |
| MeSH Terms: | Electroencephalography*/methods , Attention*/physiology , Visual Perception*/physiology , Multilayer Perceptrons* , Brain-Computer Interfaces*, Humans ; Signal Processing, Computer-Assisted |
| Abstract: | Rapid serial visual presentation (RSVP) enables efficient electroencephalography (EEG)-based brain-computer interfaces, yet single-trial decoding remains difficult due to signal overlap and multicomponent entanglement. This work developed DisCo-Former, a Transformer-based framework incorporating three priors-guided components, including trend-periodicity disentanglement, channel-level embeddings that preserve global temporal pattern, and contrastive learning that exploits target-adjacent nontargets. Although DisCo-Former surpassed existing approaches, analysis revealed a consistent attention collapse: attention maps became nearly uniform, and value projection weights shrank toward zero. Removing the Transformer encoder yields DisCo-MLP, a purely multilayer perceptron (MLP) variant that preserves all remaining modules. Across two datasets and three evaluation regimes, DisCo-MLP matched or outperformed its Transformer-based counterpart. In within-subject decoding, mean AUCs ranged from approximately 0.94 to 0.98 across two datasets, consistently exceeding strong baselines. These results indicate that, for RSVP-EEG decoding, effectiveness stems less from architectural complexity and more from modeling the signal's structure. Simplicity motivated by paradigm-specific neurophysiological priors offers a practical path to state-of-the-art performance in EEG-based interfaces. |
| Contributed Indexing: | Keywords: Brain–computer interface (BCI); attention collapse; electroencephalography (EEG); multilayer perceptron (MLP); rapid serial visual presentation (RSVP) |
| Entry Date(s): | Date Created: 20260410 Date Completed: 20260714 Latest Revision: 20260714 |
| Update Code: | 20260715 |
| DOI: | 10.1142/S0129065726500309 |
| PMID: | 41958205 |
| Database: | MEDLINE |
Be the first to leave a comment!