Application of convolutional neural networks based on the kolmogorov–arnold architecture with a wavelet activation function for identifying cognitive brain states using EEG data

Background. Analyzing cognitive states using electroencephalography (EEG) is a key task for brain-computer interfaces. Despite the success of deep learning methods, classical convolutional neural networks (CNNs) face significant limitations when processing non-stationary EEG signals: they struggle to adapt to local time-frequency features, and their decisions are difficult to interpret. Integrating the Kolmogorov–Arnold Network (KAN) architecture with wavelet transformations opens up new possibilities for improving the accuracy and interpretability of EEG analysis models by replacing linear weights with learnable nonlinear functions.

Aim. To study the impact of replacing conventional convolutional layers with layers based on the Kolmogorov–Arnold architecture utilizing various wavelet activation functions on the quality of identifying cognitive states of varying complexity.

Methods. Data from the open-access COG-BCI dataset were used, comprising recordings from 29 participants performing cognitive tasks N-Back (working memory) and MATB-II (multitasking). The original FFCLNetEEG architecture was modified by replacing CNN layers with KAN and WavKAN convolutional layers using various wavelets (Gabor, DOG, Mexican Hat, Morlet, Meyer, Shannon). Evaluation was performed using Accuracy, Precision, Recall, F1-score, and Specificity metrics with pairwise Wilcoxon tests with Bonferroni correction (p < 0,05).

Results. In the N-Back task, the FFCLNetEEG-WavKAN-DOG model showed the best performance achieving a mean accuracy of 0.790±0.059; this significantly outperformed FFCLNetEEG-KAN, FFCLNetEEG-WavKAN-Meyer and FFCLNetEEG-WavKAN-Morlet (p < 0,001 after correction). In the MATB-II task, the best performance was achieved by the FFCLNetEEGWavKAN-Gabor model performed best (mean accuracy 0.892±0.005). Analysis of three independent sessions showed high reproducibility of the results with correlation coefficients between sessions up to 0.653 for the best-performing models.

Conclusion. Integrating the Kolmogorov–Arnold architecture with embedded wavelet functions enables a reliable improvement in the classification quality of cognitive states based on EEG data. The choice of the optimal wavelet depends on the type of cognitive task: DOG for working memory (N-Back) and Gabor for multitasking (MATB-II). The findings have practical implications for the development of neuroadaptive interfaces and systems for cognitive monitoring and rehabilitation.

1. Andrikov DA, Andrikov DmA, Berezkin DV, Popov AYU, Proletarskij AV. Nejrosetevaya grafovaya arhitektura prozrachnogo iskusstvennogo intellekta v medicine. Vrach i informacionnye tekhnologii. 2025; 2: 70-83. (In Russ.) doi: 10.25881/18110193_2025_2_70.

2. Antipov VM, Badarin AA, Kurkin SA, Kiselev AR, Hramov AE. Programmno-apparatnyj kompleks dlya reabilitacii pacientov s kognitivnymi i motornymi narusheniyami. Vrach i informacionnye tekhnologii. 2024; 4: 38-47. (In Russ.) doi: 10.25881/18110193_2024_4_38.

3. Dedkov AE, Andrikov DA, Hramov AE. Obzor sposobov izmereniya kognitivnoj nagruzki mozga i metodov mashinnogo obucheniya dlya ih identifikacii na osnove dannyh EEG. Vrach i informacionnye tekhnologii. 2024; 3: 20-31. (In Russ.) doi: 10.25881/18110193_2024_3_20.

4. Kumar UA, Dhamija Y. Comparative analysis of SOM neural network with K-means clustering algorithm. In: 2010 IEEE International Conference on Management of Innovation & Technology. IEEE; 2010: 55-59.

5. Ahmadi N, Naderi M, Ghazizadeh A, Nahvi A. Eye State Identification Utilizing EEG Signals: A Combined Method Using Self-Organizing Map and Deep Belief Network. Sci Program. 2022; 2022: 4439189. doi: 10.1155/2022/4439189.

6. Yau J, Medeiros D, Prabhu P, Tiwari P, Prabhu G. Uncovering Patterns of Brain Activity from EEG Data Consistently Associated with Cybersickness Using Neural Network Interpretability Maps. arXiv preprint arXiv: 2512.20620. 2025.

7. Zhang J, Zhang W, Wang Y, et al. Combining self-organizing mapping and supervised affinity propagation clustering approach to investigate functional brain networks involved in motor imagery and execution with fMRI measurements. Front Hum Neurosci. 2015; 9: 400. doi: 10.3389/fnhum.2015.00400.

8. Becerra-Sánchez P, Reyes-Munoz A, Guerrero-Ibañez A. Feature selection model based on EEG signals for assessing the cognitive workload in drivers. Sensors. 2020; 20(20): 5881. doi: 10.3390/s20205881.

9. Kocyigit E, Alkan E, Albayrak A, et al. Enhanced feature selection using genetic algorithm for machinelearning- based phishing URL detection. Appl Sci. 2024; 14(14): 6081. doi: 10.3390/app14146081.

10. Arif M, Li H, Khalil K, et al. A comprehensive survey of evolutionary algorithms and metaheuristics in brain EEG-based applications. J Neural Eng. 2024; 21(5): 051002. doi: 10.1088/1741-2552/ad7a5b.

11. Joutsiniemi SL, Kaski S, Larsen TA. Self-organizing map in recognition of topographic patterns of EEG spectra. IEEE Trans Biomed Eng. 1995; 42(11): 1062-1068. doi: 10.1109/10.469374.

12. Huang Z, Wang M. A review of electroencephalogram signal processing methods for brain-controlled robots. Cogn Robot. 2021; 1: 111-124. doi: 10.1016/j.cogr.2021.09.002.

13. Asvestas P, Kariotou F, Kitsos G, et al. Use of genetic algorithm for the selection of EEG features. J Phys Conf Ser. 2015; 633: 012123. doi: 10.1088/1742-6596/633/1/012123.

14. Ivaylov I. EEG classification for BCI using genetic algorithm and k-fold cross validation. Electr Eng Electron. 2021; 56(3-4): 38-45.

15. Johansson R. NSGA-II design for feature selection in EEG classification related to motor imagery. 2020.

16. Hramov AE, Koronovskii AA, Makarov VA, Maksimenko VA, Pavlov AN, Sitnikova EYu. Wavelets in Neuroscience. 2nd ed. Springer Series in Synergetics. 2021. doi: 10.1007/978-3-030-75992-6.

17. Pavlov AN, Hramov AE, Koronovskii AA, Sitnikova EY, Makarov VA, Ovchinnikov AA. Wavelet analysis in neurodynamics. Phys Usp. 2012; 55(9): 845-875. doi: 10.3367/UFNe.0182.201209a.0905.

18. Murata A. An Attempt to Evaluate Mental Workload Using Wavelet Transform of EEG. Hum Factors. 2005; 47(3): 498-508. doi: 10.1518/001872005774860096.

19. Grubov VV, Maksimenko VA, Hramov AE, et al. Open-loop neuroadaptive system for enhancing student’s cognitive abilities in learning. IEEE Access. 2024; 12: 49034-49049. doi: 10.1109/ ACCESS.2024.3383847.

20. Liu Z, Wang Y, Vaidya S, et al. KAN: Kolmogorov-Arnold Networks. arXiv preprint arXiv: 2404.19756. 2024.

21. Bozorgasl Z, Chen H. Wav-KAN: Wavelet Kolmogorov–Arnold Networks. arXiv preprint arXiv:2405.12832. 2024. doi: 10.48550/arXiv.2405.12832.

22. Drokin I. Kolmogorov-Arnold Convolutions: Design Principles and Empirical Studies. arXiv preprint arXiv:2407.01092. 2024. 23. Hinss MF, Jahanpour ES, Somon B, Pluchon L, Dehais F, Roy RN. Open multi-session and multi-task EEG cognitive Dataset for passive brain-computer Interface Applications. Sci Data. 2023; 10(1): 85. doi: 10.1038/s41597-022-01898-y.

23. Li H, Chen Z, Huang Z, et al. Motor imagery EEG classification algorithm based on CNN-LSTM feature fusion network. Biomed Signal Process Control. 2022; 72: 103342. doi: 10.1016/j.bspc.2021.103342.

24. Pulver D, Angkan P, Hungler P, Etemad A. EEG-based Cognitive Load Classification using Feature Masked Autoencoding and Emotion Transfer Learning. In Proceedings of the 2023 International Conference on Multimodal Interaction ICMI '23. Р.1-8.

25. Li Z, Zhang R, Zeng Y, Tong L, Lu R, Yan B. MST-net: A multi-scale swin transformer network for EEGbased cognitive load assessment. Research report. doi: 10.1016/j.ijhcs.2023.103259.

26. Li Y, Li K, Wang S, Wu H, Li P. A spatiotemporal separable graph convolutional network for oddball paradigm classification under different cognitive-load scenarios. Expert Systems With Applications. 2024; 262: 125303. doi: 10.1016/j.eswa.2024.125303.

27. Yoo G, Kim H, Hong S. Prediction of Cognitive Load from Electroencephalography Signals Using Long Short-Term Memory Network. Bioengineering. 2023; 10: 361. doi: 10.3390/bioengineering10030361.

28. Yedukondalu J, Sharma D, Sharma LD. Subject-Wise Cognitive Load Detection Using Time-Frequency EEG and Bi-LSTM. Arab J Sci Eng. doi: 10.1007/s13369-023-08494-1.

29. Yedukondalu, J., Sunkara, K., Radhika, V. et al. Cognitive load detection through EEG lead wise feature optimization and ensemble classification. Sci Rep (2024). Статья в открытом доступе, представлен- ная в анализируемом файле s41598-024-84429-6.pdf.

30. Yedukondalu, J., Sharma, L.D. Cognitive load detection using circulant singular spectrum analysis and Binary Harris Hawks Optimization based feature selection. (Журнал не указан в предоставленном фрагменте, но для ссылки можно использовать заголовок и авторов).