Norm | Formula | Description |
---|---|---|
Chebyshev | \(d_{i,j} = \max _{p} |s_{i,p} - s_{j,p}|\) | Metric induced by the supremum norm |
Euclidean | \(d_{i,j} = \sqrt{\sum _{p=1}^{m}(s_{i,p} - s_{j,p})^2}\) | Length of a line segment between two vectors |
Manhattan | \(d_{i,j} = \sum _{p=1}^{m}|s_{i,p} - s_{j,p}|\) | Sum of the lengths of the projections onto axes |
Correlation | \(d_{i,j} = 1 - \frac{(s_i-\overline{s_i})\cdot (s_j-\overline{s_j})}{\sqrt{\sum _{p=1}^{m}s_{i,p}^2 \cdot \sum _{p=1}^{m}s_{i,p}^2 }}\) | Correlation distance between two vectors |
Cosine | \(d_{i,j} = 1 - \frac{(s_i,s_j)}{\sqrt{\sum _{p=1}^{m}s_{i,p}^2 \cdot \sum _{p=1}^{m}s_{i,p}^2 }}\) | Cosine distance between two vectors |