Module dimensionality_reducer 

Dimensionality Reducer

PCA-based dimensionality reduction for embedding vectors. Implements power-iteration PCA for top-k eigenvector extraction, plus a thin Truncated SVD wrapper.

Algorithm

1. Center the data (subtract column means).
2. Compute the covariance matrix.
3. Use power iteration with deflation to extract the top-k eigenvectors.
4. Project data onto the component subspace.

Structs

PcaReducer

PCA dimensionality reducer using power-iteration eigenvector extraction.

ReductionResult

Output of a fit_transform call.

TruncatedSvd

Truncated SVD: operates on the raw (uncentered) data matrix.

Enums

ReductionError

Errors from dimensionality reduction operations.

Functions

center_data

Center the data by subtracting column means. Returns (centered_data, column_means).

dot_product

Compute the dot product of two equal-length slices.

mat_vec_mul

Multiply matrix mat (m × n) by vector vec (n) → result (m).

normalize

Normalise a vector in-place (L2 norm). No-op for zero vectors.