ICA
Independent Component Analysis separates multivariate signals into maximally independent components
ICA
Independent Component Analysis separates multivariate signals into maximally independent components.
When to use:
- Blind source separation (e.g., cocktail party problem)
- Remove artifacts from signals (EEG, fMRI)
- Feature extraction for audio
- When features should be independent (not just uncorrelated like PCA)
Strengths: Finds independent sources, good for mixed signals, non-Gaussian assumptions Weaknesses: Sensitive to initialization, assumes linear mixing, cannot determine component order
Model Parameters
N Components (default: 2, required) Number of independent components to extract.
Algorithm (default: "parallel") ICA algorithm:
- parallel: Faster, extracts all components simultaneously (default)
- deflation: Slower, extracts one component at a time
Max Iterations (default: 200) Maximum iterations for convergence.
- 200: Usually sufficient
- 500+: For difficult data
Tolerance (default: 0.0001) Convergence threshold.
Random State (default: 42) Seed for reproducibility.