Abstract:Electroencephalography (EEG) is widely researched for neural decoding in Brain Computer Interfaces (BCIs) as it is non-invasive, portable, and economical. However, EEG signals suffer from inter- and intra-subject variability, leading to poor performance. Recent technological advancements have led to deep learning (DL) models that have achieved high performance in various fields. However, such large models are compute- and resource-intensive and are a bottleneck for real-time neural decoding. Data distribution shift can be handled with the help of domain adaptation techniques of transfer learning (fine-tuning) and adversarial training that requires model parameter updates according to the target domain. One such recent technique is Parameter-efficient fine-tuning (PEFT), which requires only a small fraction of the total trainable parameters compared to fine-tuning the whole model. Therefore, we explored PEFT methods for adapting EEG-based mental imagery tasks. We considered two mental imagery tasks: speech imagery and motor imagery, as both of these tasks are instrumental in post-stroke neuro-rehabilitation. We proposed a novel ensemble of weight-decomposed low-rank adaptation methods, EDoRA, for parameter-efficient mental imagery task adaptation through EEG signal classification. The performance of the proposed PEFT method is validated on two publicly available datasets, one speech imagery, and the other motor imagery dataset. In extensive experiments and analysis, the proposed method has performed better than full fine-tune and state-of-the-art PEFT methods for mental imagery EEG classification.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve the problem of poor EEG signal classification performance in electroencephalogram (EEG) - based brain - computer interfaces (BCIs) due to inter - individual and intra - individual differences. Specifically, the paper focuses on how to improve the classification performance of EEG signals in mental imagery tasks through parameter - efficient fine - tuning (PEFT) methods, especially for speech imagery (SI) and motor imagery (MI) tasks. #### Main problems: 1. **Inter - individual and intra - individual differences in EEG signals**: EEG signals are easily affected by individual differences, resulting in unstable performance of the model on different individuals or at different time periods of the same individual. 2. **Computational resource requirements of deep - learning models**: Although traditional deep - learning models perform well in many fields, they usually have a large number of parameters and high computational complexity, making it difficult to achieve real - time neural decoding. 3. **Domain adaptation problem**: When the data distribution changes (for example, from one individual to another), effective domain adaptation techniques are required to reduce the impact of this distribution change on the model performance. #### Proposed solutions: - **Parameter - efficient fine - tuning (PEFT)**: By introducing methods such as low - rank adapters, only a small number of parameters are updated, thereby achieving efficient and low - resource - consuming model fine - tuning. - **EDoRA method**: A novel method of ensemble of weight - decomposed low - rank adapters (EDoRA) is proposed to improve the classification performance of EEG signals in mental imagery tasks. - **Cross - task adaptation**: It has been verified that the pre - trained model parameters of one mental imagery task are used for the fine - tuning of another mental imagery task, and good results have been achieved. Through these methods, the paper hopes to significantly reduce the computational resource requirements while maintaining high performance, making EEG signal classification more suitable for practical applications, especially in rehabilitation treatment in the field of brain - computer interfaces.

EEG-Based Mental Imagery Task Adaptation via Ensemble of Weight-Decomposed Low-Rank Adapters

EEG-Based Motor Imagery Recognition Framework via Multisubject Dynamic Transfer and Iterative Self-Training

Calibration-free online test-time adaptation for electroencephalography motor imagery decoding

A Dynamic Domain Adaptation Deep Learning Network for EEG-based Motor Imagery Classification

Optimal Deep Learning-Based Recognition Model for EEG Enabled Brain-Computer Interfaces Using Motor-Imagery

Adaptive transfer learning for EEG motor imagery classification with deep Convolutional Neural Network

Session-independent subject-adaptive mental imagery BCI using selective filter-bank adaptive Riemannian features

EEG2IMAGE: Image Reconstruction from EEG Brain Signals

Empowering EEG motor imagery classification with deep transfer learning approach

Feature Reweighting for EEG-based Motor Imagery Classification

Learning EEG Representations with Weighted Convolutional Siamese Network: A Large Multi-Session Post-Stroke Rehabilitation Study.

Improved Domain Adaptation Network Based on Wasserstein Distance for Motor Imagery EEG Classification

Multiattention Adaptation Network for Motor Imagery Recognition

Physics-informed and Unsupervised Riemannian Domain Adaptation for Machine Learning on Heterogeneous EEG Datasets

Visual Decoding and Reconstruction via EEG Embeddings with Guided Diffusion

Optimization enabled deep residual neural network for motor imagery EEG signal classification

Subject-Adaptive Transfer Learning Using Resting State EEG Signals for Cross-Subject EEG Motor Imagery Classification

[Value of serological diagnosis of celiac disease for the determination of intolerance to prolamines of certain varieties of oats in patients with celiac disease].

On-device Learning of EEGNet-based Network For Wearable Motor Imagery Brain-Computer Interface

Latent Alignment with Deep Set EEG Decoders

Motor Imagery Decoding Using Ensemble Curriculum Learning and Collaborative Training