HyperdimensionalComputingForEpilepsy

Repository structure

Most of the code is related to work of Una Pale during her PhD thesis on the topic "Hypersimensional computing for biosignal monitoring: Applications for epilepsy detection" Each subfolder is related to one subtopic and chapter of the thesis.

01_DiffFeaturesForEpilepsy

Testing different features in combination with HD computing for epileptic seizures detection.

Paper: „Systematic Assessment of Hyperdimensional Computing for Epileptic Seizure Detection“, Una Pale, Tomas Teijeiro, David Atienza

• 2021 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC),https://ieeexplore.ieee.org/document/9629648
• Arxiv version: https://arxiv.org/abs/2105.00934

02_FeatureSelectionWithHD

Proposing different encoding approaches for spatio-temporal data like EEG, and using it for feature selection as part of HD workflow. Tested for epileptic seizures detection.

Paper: „ExG Signal Feature Selection Using Hyperdimensional Computing Encoding“, Una Pale, Tomas Teijeiro, David Atienza

• 2022 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), https://ieeexplore.ieee.org/abstract/document/9995107
• Arxiv version: https://arxiv.org/abs/2205.07654

03_MultiCentroidLearning

Instead of allowing only one hypervectors model per class, complexity of epileptic seizure signatures inspired us to propose smarter learning approache.

Paper: „Multi-Centroid Hyperdimensional Computing Approach for Epileptic Seizure Detection“, Una Pale, Tomas Teijeiro, David Atienza

• Frontiers in Neurology, 13, 1-13, 816294, https://www.frontiersin.org/articles/10.3389/fneur.2022.816294/full
• Arxiv version: https://arxiv.org/abs/2111.08463

04_LearningImprovements

We systematically compare multi-centroid and other proposed 'smarter' learning approaches in the literature of HD computing (iterative learning and online learning). Paper: „Exploration of Hyperdimensional Computing Strategies for Enhanced Learning on Epileptic Seizure Detection“, Una Pale, Tomas Teijeiro, David Atienza

• 2022 44th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), https://ieeexplore.ieee.org/document/9870919
• Arxiv version: https://arxiv.org/abs/2201.09759

05_PersAndGenModels

Testing interplay between personal and general models for epilepsy detection and proposing new idea of hybrid models that HD computing allows. We also propose how this models can be used for knowledge transfer between different epilepsy datasets.

Paper: „Combining General and Personalized Models for Epilepsy Detection with Hyperdimensional Computing“, Una Pale, Tomas Teijeiro, David Atienza

• Submitted to 2023 Conference on Lifelong Learning Agents (CoLLAs)
• Arxiv version: https://arxiv.org/abs/2303.14745

06_Explainability

Demonstrating how HD computing and the way encoding is designed can be used to backtrack influence of individual features and channels for seizure detection in time.