MDS
Multidimensional Scaling preserves pairwise distances between points in lower dimensions
MDS
Multidimensional Scaling preserves pairwise distances between points in lower dimensions.
When to use:
- Have distance/dissimilarity matrix
- Want to preserve all pairwise distances
- Need visualization of relationships
- Small to medium datasets
Strengths: Preserves distances well, works with any distance matrix, interpretable Weaknesses: Very slow (O(n³)), no inference on new data, sensitive to outliers
Model Parameters
N Components (default: 2, required) Embedding dimensions.
Metric (default: true) Type of MDS:
- true: Metric MDS - preserves actual distances (default)
- false: Non-metric MDS - preserves rank order only
Max Iterations (default: 300) Maximum optimization iterations.
- 300: Standard (default)
- 500-1000: Better convergence for difficult data
Random State (default: 42) Seed for reproducibility.