LUNA: Efficient and Topology-Agnostic Foundation Model for EEG Signal Analysis
abnormality detectionartifact rejectioncross-attentionelectrode geometrieselectroencephalographyemotion recognitionlatent unified network architecturelearned querieslunamasked-patch reconstructionmulti-channel eegpatch-wise temporal self-attentionself-supervised foundation modeltopological heterogeneitytopology-agnostic latent space
Electroencephalography (EEG) offers a non-invasive lens into human brain activity, but building large‐scale models is hampered by $\textit{topological heterogeneity}$: each public corpus defines its own electrode layout, limiting generalization. We introduce $\textbf{LUNA}$ ($\textbf{L}$atent $\textbf{U}$nified $\textbf{N}$etwork $\textbf{A}$rchitecture), a self-supervised foundation model that reconciles disparate electrode geometries while scaling linearly