dimensionality-reduction.md

Dimensionality Reduction

Reduce vector dimensions while preserving important information using PCA and whitening.

PCA (Principal Component Analysis)

Reduce dimensions while preserving variance:

-- PCA transformation
SELECT pca_transform(
    'data_table',
    'features',
    128,  -- target dimensions
    'pca_model'
);

-- Apply PCA to new data
SELECT pca_apply(features, 'pca_model') AS reduced_features
FROM test_table;

PCA Whitening

Standardize variance across components:

-- PCA with whitening
SELECT pca_whiten(
    'data_table',
    'features',
    128,
    'pca_whitened_model'
);

Benefits

• Reduce storage requirements
• Speed up training and inference
• Remove noise and redundant information
• Visualize high-dimensional data

Learn More

For detailed documentation on PCA, whitening, choosing dimensions, and inverse transformation, visit:

Dimensionality Reduction Documentation

Related Topics

• Clustering - Apply clustering after reduction
• Quality Metrics - Evaluate reduction quality