Automatic emotion recognition based on EEG has been a key research frontier in recent years, involving the direct extraction of emotional states from brain dynamics. However, existing deep learning approaches often treat EEG either as a sequence or as a static spatial map, thereby failing to jointly capture the temporal evolution and spatial dependencies underlying emotional responses. To address this limitation, we propose an Interpretable Residual Spatio-Temporal Graph Attention Network (IRSTGANet) that integrates temporal convolutional encoding with residual graph-attention blocks. The temporal module enhances short-term EEG dynamics, while the graph-attention layers learn adaptive node connectivity relationships and preserve contextual information through residual links. Evaluated on the DEAP and SEED datasets, the proposed model achieved exceptional performance on valence and arousal, as well as four-class and nine-class classification on the DEAP dataset and on the three-class SEED dataset, exceeding state-of-the-art methods. These results demonstrate that combining temporal enhancement with residual graph attention yields both improved recognition performance and interpretable insights into emotion-related neural connectivity.

Spatiotemporal deep learning models, like Graph Neural Networks (GNNs), Transformers, and selective State Space Models (Mamba), have achieved impressive performance in electroencephalogram (EEG) decoding and affective computing. However, their generalization performance often degrades severely under subject-independent Leave-One-Subject-Out (LOSO) cross-validation protocols. This generalization drop is often attributed to generic domain shifts in standard Brain–Computer Interface (BCI) literature, and is tackled by parameter-heavy adaptations. In contrast, this paper proposes a unified mathematical diagnostic framework to audit and measure the underlying representation and geometric collapse in spatiotemporal brain–computer interfaces. More concretely, we formalize: (1) Topological over-smoothing under volume conduction through Graph Dirichlet Energy bounds indicating GCNs as low-pass filters that smooth localized electrode variations; (2) representation collapse through the Normalized Rank Uniformity Index (NRUI) based on the Shannon Entropy of latent covariance eigenvalues, that distinguishes between dimensional and semantic collapse; and (3) geometric manifold distortions under subject domain shifts on the Symmetric Positive Definite (SPD) Riemannian manifold under the Affine-Invariant Riemannian Metric (AIRM) projection. Auditing these diagnostic metrics on canonical models across DEAP, DREAMER, and SEED, we demonstrate why standard spatiotemporal architectures suffer from performance collapse in cross-subject configurations. We provide BCI engineers with a tangible mathematical blueprint to design robust, collapse-resistant decoders.

Self-supervised graph collaborative filtering suffers from two geometric pathologies. In Dimensional Collapse, embeddings quietly collapse into a low-rank subspace, a well-documented but poorly solved problem. Semantic Collapse is more complex: stiff $L^2$-squared orthogonalization penalties that are supposed to push the embeddings apart end up ripping through the heavy-tailed community overlaps that contain the collaborative signal. Earlier studies attempted to mitigate sparsity by injecting static noise or structural perturbations, but such interventions did not pinpoint the root cause, i.e., the distortions in the global covariance geometry itself. In this paper, we propose a Huber-Contrastive Graph Convolutional Network (HCGCN) that combines a spatial message-passing backbone with an $O(1)$ contrastive augmentation overhead and a Relation-Aware Dual-View Gated Contrastive Network. The main novelty is a Huber Covariance Whitening module that imposes a geometry-aware threshold on the cross-correlation matrix of augmented views—below the threshold, the gradients follow an $L^2$ penalty (enforcing uniformity); above it, the penalty flattens to $L^1$ (protecting genuine semantic clusters from gradient explosion). This theoretically motivated dual-regime penalty actively preserves the macro-semantic topology of the graph while aggressively stamping out spurious noise correlations. The HCGCN is evaluated on the Yelp2018, Amazon-Book, and MovieLens datasets and performs significantly better than state-of-the-art baselines like LightGCN, SGL, SimGCL, and NESCL, especially under severe cold-start settings where covariance regulation proves most critical.