Agglomerative hierarchical clustering ", - "initially assumes that all time series are forming their own clusters. It then grows a clustering dendrogram thanks to 2 inputs:

",
- "First, a **dissimilarity matrix** between all pairs ",
+ "initially assumes that all time series are forming their own clusters. It then grows a clustering dendrogram using two inputs:

",
+ "A **dissimilarity matrix** between all pairs ",
"of time series is calculated with one of the metrics, such as ",
- "Euclidean (L2 norm) ",
- "or Manhattan (L1 norm) distance. ",
- "Dynamic Time Warping (DTW) ",
- "is another distance metric that does not only compare series point by point but also tries to align them such that shapes between the 2 series are matched. ",
+ "Euclidean (L2 norm), ",
+ "Manhattan (L1 norm), or ",
+ "Dynamic Time Warping (DTW). ",
+ "Instead of comparing time series point by point, DTW tries to align and match their shapes. ",
"This makes DTW a good quantification of similarity when signals are similar but shifted in time.

In the second step, clusters are successively built and merged together. The distance between the newly formed clusters is determined by the **linkage criterion** ",
"using one of linkage methods.