Module dimensionality 

Functions

binned_centred 🔒

demixed_from_centered

Demixed PCA of a grand-mean-centred condition-mean matrix (n_cond × t). Returns (projected [n_cond × nc], explained_variance_ratio [nc]).

demixed_pca

Demixed PCA. Kobak et al. 2016. conditions: list of (condition trains).

fa_from_centered

Factor-analysis EM of a mean-centred matrix mat (d × t).

factor_analysis

Factor analysis via EM. Rubin & Thayer 1982.

pca_from_centered

PCA of a mean-centred matrix mat (d × t, row-major). Returns (projected [nc × t], explained_variance_ratio [nc]).

spd_inverse 🔒

SPD inverse via Cholesky (needed for the explicit nf · M⁻¹ term).

spd_solve 🔒

Solve the SPD system A X = B via Cholesky. a is row-major n × n, b is row-major n × k; returns X row-major n × k. A is never inverted.

spike_train_pca

PCA on binned spike trains. trains: list of binary trains.

symmetric_eigen 🔒

Descending eigenvalues and sign-canonicalised eigenvectors of a symmetric matrix a (row-major n × n) via nalgebra’s symmetric eigensolver (tridiagonalisation + implicit QR — LAPACK-grade, replacing a hand-rolled Jacobi sweep). Each eigenvector column is sign-fixed so its largest-magnitude entry is positive, making downstream projections deterministic across backends. Eigenvectors are returned row-major (vecs[row * n + col]).