Describe the feature or idea you want to propose
Recent advances in similarity-based classification, like the Proximity Forest 2.0 include parameterising cost functions used in similarity measures. However, the current aeon implementation of dtw uses squared distance as the cost function. The Minkowski distance is a generalised form of both the Euclidean and Manhattan distance, and hence can be used as the cost function for dtw and adtw.
Describe your proposed solution
Use Minkowski distance instead of squared_univariate_distance to compute the dtw cost matrix.
distance = np.sum((np.abs(x - y) ** p)) ** (1.0 / p)
Describe alternatives you've considered, if relevant
No response
Additional context
No response
Describe the feature or idea you want to propose
Recent advances in similarity-based classification, like the Proximity Forest 2.0 include parameterising cost functions used in similarity measures. However, the current aeon implementation of
dtwuses squared distance as the cost function. The Minkowski distance is a generalised form of both the Euclidean and Manhattan distance, and hence can be used as the cost function fordtwandadtw.Describe your proposed solution
Use Minkowski distance instead of
squared_univariate_distanceto compute thedtwcost matrix.distance = np.sum((np.abs(x - y) ** p)) ** (1.0 / p)Describe alternatives you've considered, if relevant
No response
Additional context
No response