
\[sim_f(t_{1},t_{2})=\frac{\sum_{k=1}^{L}f(t_{1}).P_{k}.f(t_{2}).P_{k}}{\sqrt{\sum_{k=1}^{L}f(t_{1}).P_{k}^{2}}.\sqrt{\sum_{k=1}^{L}f(t_{2}).P_{k}^{2}}}\]
1. Start from the user specified feature
2. Search in neighborhood of existing pertinent features
3. Expand search range gradually
\[ sim(t_{1},t_{2})=\sum_{i=1}^{L}sim_{f_{i}}(t_{1},t_{2}).f_{i}.weight \]