tensor_classification

tensor_classification performs tensor-based classification by applying multilinear discriminant analysis and tensor projection methods to extract discriminatory features from multilinear datasets such as EEG and BCI recordings.

Key Features:

• Multilinear Discriminant Analysis (MDA): Implements supervised feature extraction for tensor-structured data using rigorously optimized MDA.
• PARAFAC and Tucker integration: Incorporates MDA with both PARAFAC and traditional Tucker tensor structures to support multiple multilinear decompositions.
• Manifold optimization: Performs optimization over the cross-product of Stiefel manifolds to provide a rigorous, monotonically improving optimization framework for MDA.
• Comparison capabilities: Enables direct comparison between proposed MDA approaches and existing methods, including unsupervised multilinear decompositions.
• Simulated data generation: Provides functionality to simulate data for methodological comparisons and validation.

Scientific Applications:

• Electroencephalography (EEG) data analysis: Extracts discriminatory patterns from EEG recordings with minimal preprocessing.
• Brain–computer interface (BCI): Applies MDA directly to raw BCI data to identify meaningful activity patterns comparable to EEG paradigms.

Methodology:

Supervised feature extraction via MDA, optimization on the cross-product of Stiefel manifolds, and simulated data generation for method comparison.