Domain-separated EEG classifiers decode familiarity within each stimulus category, with peak LOGO Mean Recall of 0.8992 for faces-only (HistGradient Boosting) and 0.8670 for equations-only (CNN)¶

Claim · All claims
Evidence · 1 study · associational r1 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? Under Leave-One-Group-Out validation, binary classifiers achieved high average single-class recall for equations-only familiarity (CNN 0.8670 ± 0.2392) and faces-only familiarity (HistGradient Boosting 0.8992 ± 0.0850), plus domain-only classification at 0.9491 (CNN). → Isuru Nanayakkara and Thilina Halloluwa 2026

Evidence¶

Isuru Nanayakkara and Thilina Halloluwa 2026¶

Isuru Nanayakkara and Thilina Halloluwa. (2026). Automating Learner Assessment: Benchmarking Machine Learning and Deep Learning Models for EEG-Based Familiarity Prediction. https://arxiv.org/abs/2608.16541 doi:10.48550/arxiv.2608.16541

q2 · i? · associational · r1

Secondary domain-isolated LOGO evaluations on 37 trial blocks (24 equation, 13 face); the article reports the faces-only HistGradient Boosting model "outperforming the CNN Optimized model". LOGO Mean Recall is a recall-based metric not directly comparable to balanced F1.

"For the equations-only familiarity task, the CNN Optimized model achieved an average LOGO Mean Recall of 0.8670 ± 0.2392, and the HistGradient Boosting classifier achieved 0.7588 ± 0.3256. For the faces-only familiarity task, the HistGradient Boosting classifier achieved a peak average LOGO Mean Recall of 0.8992 ± 0.0850, outperforming the CNN Optimized model which achieved 0.8875 ± 0.2347."

Discussion¶

Related Claims¶

• Trial-level spectral features robustly differentiate stimulus domains (faces vs. equations) but no familiarity-driven spectral shift survives FDR correction — related
• Trial-independent EEG familiarity classification remains statistically significantly above chance for the 4-class, equations-only, and faces-only tasks under Bonferroni-corrected permutation tests — related
• Temporal leakage from shuffled neighboring epochs inflates EEG familiarity classification, with peak weighted F1 dropping from 0.9853 (CNN, stratified) to 0.6038 (CNN, Group K-Fold) — related
• Familiarity recognition is represented by a distributed multivariate spatial-spectral pattern rather than localized single-feature shifts, necessitating multivariate machine learning — related
• Offloading is decodable from metacognitive monitoring and control brain processes — related
• Random Forest and SHAP analyses identify Gamma and Beta band power at temporal and frontal channels (T3, P3, F7, T4, T4, F8, F7) as the most critical features for EEG familiarity classification — related