Metastable Substructure Embedding and Robust Classification of Multichannel EEG Data Using Spectral Graph Kernels

Future Internet, Journal Year: 2025, Volume and Issue: 17(3), P. 102 - 102

Published: Feb. 23, 2025

Classification of neurocognitive states from Electroencephalography (EEG) data is complex due to inherent challenges such as noise, non-stationarity, non-linearity, and the high-dimensional sparse nature connectivity patterns. Graph-theoretical approaches provide a powerful framework for analysing latent state dynamics using measures across spatio-temporal-spectral dimensions. This study applies graph Koopman embedding kernels (GKKE) method extract neuro-markers seizures epileptiform EEG activity. EEG-derived graphs were constructed correlation mean phase locking value (mPLV), with adjacency matrices generated via threshold-binarised connectivity. Graph kernels, including Random Walk, Weisfeiler–Lehman (WL), spectral-decomposition (SD) evaluated space feature extraction by approximating spectral decomposition. The potential embeddings in identifying metastable structures has been demonstrated empirical analyses. robustness these features was classifiers Decision Trees, Support Vector Machine (SVM), Forest, on Epilepsy-EEG Children’s Hospital Boston’s (CHB)-MIT dataset cognitive-load-EEG datasets online repositories. classification workflow combining mPLV measure, WL kernel, Tree (DT) outperformed alternative combinations, particularly considering accuracy (91.7%) F1-score (88.9%), comparative investigation presented results section convinces that employing cost-sensitive learning improved mPLV-WL-DT 91% compared 88.9% without learning. work advances EEG-based neuro-marker estimation, facilitating reliable assistive tools prognosis cognitive training protocols.

Language: Английский

---

RandONets: Shallow Networks with Random Projections for Learning Linear and Nonlinear Operators

Journal of Computational Physics, Journal Year: 2024, Volume and Issue: unknown, P. 113433 - 113433

Published: Oct. 1, 2024

Language: Английский

---

Early warning indicators via latent stochastic dynamical systems

Chaos An Interdisciplinary Journal of Nonlinear Science, Journal Year: 2024, Volume and Issue: 34(3)

Published: March 1, 2024

Detecting early warning indicators for abrupt dynamical transitions in complex systems or high-dimensional observation data are essential many real-world applications, such as brain diseases, natural disasters, and engineering reliability. To this end, we develop a novel approach: the directed anisotropic diffusion map that captures latent evolutionary dynamics low-dimensional manifold. Then three effective signals (Onsager–Machlup indicator, sample entropy transition probability indicator) derived through coordinates stochastic systems. validate our framework, apply methodology to authentic electroencephalogram data. We find capable of detecting tipping point during state transition. This framework not only bridges with but also shows potential ability automatic labeling on time series.

Language: Английский

---

Advancing Colorectal Cancer Diagnosis with AI-Powered Breathomics: Navigating Challenges and Future Directions

Diagnostics, Journal Year: 2023, Volume and Issue: 13(24), P. 3673 - 3673

Published: Dec. 15, 2023

Early detection of colorectal cancer is crucial for improving outcomes and reducing mortality. While there strong evidence effectiveness, currently adopted screening methods present several shortcomings which negatively impact the early stage carcinogenesis, including low uptake due to patient discomfort. As a result, developing novel, non-invasive alternatives an important research priority. Recent advancements in field breathomics, study breath composition analysis, have paved way new avenues effective monitoring. Harnessing utility Volatile Organic Compounds exhaled breath, breathomics has potential disrupt practices. Our goal outline key efforts this area focusing on machine learning used analysis data, highlight challenges involved artificial intelligence application context, suggest possible future directions are considered within framework European project ONCOSCREEN.

Language: Английский

---

Learning the latent dynamics of fluid flows from high-fidelity numerical simulations using parsimonious diffusion maps

Physics of Fluids, Journal Year: 2024, Volume and Issue: 36(10)

Published: Oct. 1, 2024

We use parsimonious diffusion maps (PDMs) to discover the latent dynamics of high-fidelity Navier–Stokes simulations with a focus on two-dimensional (2D) fluidic pinball problem. By varying Reynolds number Re, different flow regimes emerge, ranging from steady symmetric flows quasi-periodic asymmetric and chaos. The proposed non-linear manifold learning scheme identifies in crisp manner expected intrinsic dimension underlying emerging over parameter space. In particular, PDMs estimate that emergent oscillatory regime can be captured by just two variables, while chaotic regime, dominant modes are three as anticipated normal form theory. On other hand, proper orthogonal decomposition/principal component analysis (POD/PCA), most commonly used for dimensionality reduction fluid mechanics, does not provide such separation between modes. To validate performance PDMs, we also compute reconstruction error, constructing decoder using geometric harmonics (GHs). show outperforms POD/PCA whole Re range. Thus, believe will allow development more accurate reduced order models simulators, relaxing curse numerical tasks bifurcation analysis, optimization, control.

Language: Английский

---

Slow Invariant Manifolds of Fast-Slow Systems of ODEs with Physics-Informed Neural Networks

SIAM Journal on Applied Dynamical Systems, Journal Year: 2024, Volume and Issue: 23(4), P. 3077 - 3122

Published: Dec. 12, 2024

.We present a physics-informed neural network (PINN) approach using symbolic differentiation for the discovery of slow invariant manifolds (SIMs), general class fast-slow dynamical systems ODEs. In contrast to other machine learning approaches that construct reduced order black-box surrogate models simple regression, and/or require priori knowledge fast and variables per se, our simultaneously decomposes vector field into components provides functional underlying SIM in closed form. The decomposition is achieved by finding transformation state ones, which enables derivation an explicit, terms variables, functional. latter obtained solving PDE corresponding invariance equation within geometric singular perturbation theory (GSPT) single-layer feedforward with differentiation. performance proposed numerical framework assessed via three benchmark problems. We also provide comparison GSPT methods, namely quasi steady approximation (QSSA), partial equilibrium (PEA), computational (CSP) one two iterations. show PINN scheme approximations equivalent or even higher accuracy than those provided QSSA, PEA, CSP, especially close boundaries SIMs.Keywordsslow manifoldsfast-slow systemsmachine learningnumerical methodsphysics-informed networksMSC codes65P9968T2034E1570K7034E13

Language: Английский

---

Early warning indicators via latent stochastic dynamical systems

arXiv (Cornell University), Journal Year: 2023, Volume and Issue: unknown

Published: Jan. 1, 2023

Detecting early warning indicators for abrupt dynamical transitions in complex systems or high-dimensional observation data is essential many real-world applications, such as brain diseases, natural disasters, and engineering reliability. To this end, we develop a novel approach: the directed anisotropic diffusion map that captures latent evolutionary dynamics low-dimensional manifold. Then three effective signals (Onsager-Machlup Indicator, Sample Entropy Transition Probability Indicator) are derived through coordinates stochastic systems. validate our framework, apply methodology to authentic electroencephalogram (EEG) data. We find capable of detecting tipping point during state transition. This framework not only bridges with but also shows potential ability automatic labeling on time series.

Language: Английский

---

Early Warning Via Transitions in Latent Stochastic Dynamical Systems

Published: Jan. 1, 2023

Early warnings for dynamical transitions in complex systems or high-dimensional observation data are essential many real world applications, such as gene mutation, brain diseases, natural disasters, financial crises, and engineering reliability. To effectively extract early warning signals, we develop a novel approach: the directed anisotropic diffusion map that captures latent evolutionary dynamics low-dimensional manifold. Applying methodology to authentic electroencephalogram (EEG) data, successfully find appropriate effective coordinates, derive signals capable of detecting tipping point during state transition. Our method bridges with original dataset. The framework is validated be accurate through numerical experiments, terms density transition probability. It shown second coordinate holds meaningful information critical various evaluation metrics.

Language: Английский

---