Abstract:Due to the abundant neurophysiological information in the electroencephalogram (EEG) signal, EEG signals integrated with deep learning methods have gained substantial traction across numerous real-world tasks. However, the development of supervised learning methods based on EEG signals has been hindered by the high cost and significant label discrepancies to manually label large-scale EEG datasets. Self-supervised frameworks are adopted in vision and language fields to solve this issue, but the lack of EEG-specific theoretical foundations hampers their applicability across various tasks. To solve these challenges, this paper proposes a knowledge-driven cross-view contrastive learning framework (KDC2), which integrates neurological theory to extract effective representations from EEG with limited labels. The KDC2 method creates scalp and neural views of EEG signals, simulating the internal and external representation of brain activity. Sequentially, inter-view and cross-view contrastive learning pipelines in combination with various augmentation methods are applied to capture neural features from different views. By modeling prior neural knowledge based on homologous neural information consistency theory, the proposed method extracts invariant and complementary neural knowledge to generate combined representations. Experimental results on different downstream tasks demonstrate that our method outperforms state-of-the-art methods, highlighting the superior generalization of neural knowledge-supported EEG representations across various brain tasks.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is **how to extract effective neural representations from electroencephalogram (EEG) signals with limited labels**. Specifically, the paper focuses on: 1. **The problem of insufficient labeled data**: Traditional supervised learning methods require a large amount of labeled data to train models for extracting task - related features. However, EEG signals are complex and dynamic, and their interpretation requires specialized knowledge in neurophysiology, signal processing, and cognitive neuroscience, making it difficult and costly to label EEG data for specific tasks on a large scale. In addition, EEG signals may vary among different individuals and conditions, making it challenging to establish consistent labeling standards. 2. **Limitations of existing self - supervised methods**: Although existing self - supervised frameworks can work effectively in fields such as computer vision and natural language processing, these methods mainly explore the spatial and contextual semantics of multi - channel EEG signals while ignoring the inherent neural mechanisms and complex dynamic neural correlations of EEG signals. Due to the lack of theoretical support from prior neural knowledge, existing EEG representation methods have difficulty in extracting general neural features applicable to different downstream tasks. To solve these problems, the paper proposes a knowledge - driven cross - view contrastive learning framework (Knowledge - Driven Cross - view Contrastive Learning, KDC2), which combines neural theories and simulates the internal and external representations of brain activities by constructing scalp views and neural topology views. By designing multiple data augmentation methods and contrastive learning strategies, KDC2 can capture neural features from different views, extract invariant and complementary neural knowledge, and generate comprehensive EEG neural representations. Experimental results show that this method outperforms existing state - of - the - art methods in different downstream tasks, demonstrating its superior generalization ability in various brain tasks.

A Knowledge-Driven Cross-view Contrastive Learning for EEG Representation

Multi-view Domain-Adaptive Representation Learning for EEG-based Emotion Recognition

CLECG: A Novel Contrastive Learning Framework for Electrocardiogram Arrhythmia Classification

Eeg2vec: Self-Supervised Electroencephalographic Representation Learning

Knowledge-guided EEG Representation Learning

Multiview Contrastive Learning for Unsupervised Domain Adaptation in Brain–Computer Interfaces

Enhancing Low-Density EEG-Based Brain-Computer Interfaces with Similarity-Keeping Knowledge Distillation

Unsupervised Multi-Source Domain Adaptation Via Contrastive Learning for EEG Classification

Prototypical Contrastive Domain Adaptation Network for Nonstationary EEG Classification

Self-supervised contrastive learning for EEG-based cross-subject motor imagery recognition

Domain-Generalized EEG Classification With Category-Oriented Feature Decorrelation and Cross-View Consistency Learning

EEGFormer: Towards Transferable and Interpretable Large-Scale EEG Foundation Model

A Multi-View Deep Learning Framework for EEG Seizure Detection

Pre-Training Graph Contrastive Masked Autoencoders are Strong Distillers for EEG

DCAE: A dual conditional autoencoder framework for the reconstruction from EEG into image

A Multi-View Unified Feature Learning Network For Eeg Epileptic Seizure Detection

Structure invariance-driven collaborative contrastive network for EEG decoding

Cross-domain EEG signal classification via geometric preserving transfer discriminative dictionary learning

Decoding visual brain representations from electroencephalography through Knowledge Distillation and latent diffusion models

A joint cross-dimensional contrastive learning framework for 12-lead ECGs and its heterogeneous deployment on SoC

Auditory Attention Decoding with Task-Related Multi-View Contrastive Learning