Description

Electroencephalography (EEG) is a widely used non-invasive modality for studying brain activity across a variety of contexts, such as brain-computer interfaces, cognitive state detection, emotion recognition, neurological diagnosis, and clinical monitoring. Despite its relevance, the automated analysis of EEG signals remains a significant challenge due to several constraints, including low signal-to-noise ratio, high inter-subject and intra-subject variability, and heterogeneity in acquisition protocols.
Deep learning approaches have led to significant improvements in the performance of EEG analysis systems [Craik et al. 2019]. However, these methods generally require large volumes of annotated data to achieve robust results. Yet in many EEG applications, data acquisition is time-consuming, annotation often requires domain-specific expertise, and the data collected may be sensitive in nature. Furthermore, models trained on one group of subjects generalize poorly to new individuals, particularly when experimental conditions, sessions, or signal distributions vary. This difficulty is especially critical for personalized systems such as brain-computer interfaces, where it is necessary to reduce calibration time while maintaining a high level of performance and robustness. In this context, few-shot learning aims to learn effectively from a very small number of annotated examples [Wang et al. 2020]. This thesis proposes to explore the potential of this approach for EEG signal analysis in situations where annotated data is limited.

Skills required

M2 Computer science

Bibliography

[1] Craik, A., He, Y., & Contreras-Vidal, J. L. (2019). Deep learning for electroencephalogram (EEG) classification tasks: a review. Journal of neural engineering, 16(3), 031001.
[2] Wang, Y., Yao, Q., Kwok, J. T., & Ni, L. M. (2020). Generalizing from a few examples: A survey on few-shot learning. ACM computing surveys (csur), 53(3), 1-34.
[3] Finn, C., Abbeel, P., & Levine, S. (2017, July). Model-agnostic meta-learning for fast adaptation of deep networks. In International conference on machine learning (pp. 1126-1135). PMLR.
[4] Chen, C., Fang, H., Yang, Y., & Zhou, Y. (2025). Model-agnostic meta-learning for EEG-based inter-subject emotion recognition. Journal of Neural Engineering, 22(1), 016008.
[5] Han, J. W., Bak, S., Kim, J. M., Choi, W., Shin, D. H., Son, Y. H., & Kam, T. E. (2024). META-EEG: Meta-learning-based class-relevant EEG representation learning for zero-calibration brain–computer interfaces. Expert Systems with Applications, 238, 121986.
[6] Ng, H. W., & Guan, C. (2024). Subject-independent meta-learning framework towards optimal training of EEG-based classifiers. Neural Networks, 172, 106108.
[7] Liu, J., Tang, P., Wang, W., Ren, Y., Hou, X., Heng, P. A., ... & Li, C. (2026). A survey on inference optimization techniques for mixture of experts models. ACM Computing Surveys, 58(10), 1-37.
[8] Patacchiola, M., Turner, J., Crowley, E. J., O'Boyle, M., & Storkey, A. J. (2020). Bayesian meta-learning for the few-shot setting via deep kernels. Advances in Neural Information Processing Systems, 33, 16108-16118.
[9] Belmonte, R., Aissaoui, A., Mihoubi, S., Allaert, B., Mennesson, J., Bilasco, I. M., & Goncalves, L. (2021, June). BAREM: A multimodal dataset of individuals interacting with an e-service platform. In 2021 International Conference on Content-Based Multimedia Indexing (CBMI) (pp. 1-6). IEEE.
[10] Chaabene, S., Bouaziz, B., Boudaya, A., Hökelmann, A., Ammar, A., & Chaari, L. (2021). Convolutional neural network for drowsiness detection using EEG signals. Sensors, 21(5), 1734.

Keywords

Machine learning, meta-learning, Few-Shot learning, EEG