Decomposition

• PCA
• TruncatedSVD
• SparsePCA

How to use the Decomposition module in real projects

The Decomposition module reduces feature dimensions while preserving useful signal for visualization, denoising, and faster downstream models.

Selection checklist

1. Use PCA for dense continuous tabular data when variance preservation is the goal.
2. Use TruncatedSVD for sparse matrices such as text vectors or recommendation matrices.
3. Use SparsePCA when you need components with feature-level sparsity for interpretability.

Common implementation workflow

1. Start from a simple baseline in this module and evaluate on a holdout split.
2. Compare at least one alternative algorithm from this module before locking production defaults.
3. Pair model quality metrics with runtime constraints (latency, memory, bundle size).

Common search intents

• pca javascript
• truncated svd typescript
• dimensionality reduction browser ml

Explore algorithms in this module

• pca
• sparsePCA
• truncatedSVD

FAQ

What problem does Decomposition solve in JavaScript machine learning projects?

Decomposition helps teams implement production-ready ML workflows in browser and Node.js environments with a familiar scikit-learn-style API.

When should I choose Decomposition instead of other Decomposition algorithms?

Use Decomposition when it best matches your data shape, labeling strategy, and runtime constraints. Benchmark against at least one alternative in the same module before finalizing defaults.

Can I run Decomposition in both browser and Node.js with @kanaries/ml?

Yes. @kanaries/ml is designed for JavaScript and TypeScript runtimes across browser applications, server-side Node.js services, and edge-friendly workflows.