Table 2 Quantitative comparison of the trained weights resulting from one iteration of the timestamp training rule vs. the DTW training rule on a randomly selected observation from the 24-h outdoor temperature time series.

From: Using dynamic time warping self-organizing maps to characterize diurnal patterns in environmental exposures

Summary statistic

Timestamp Alignment

DTW Alignment

Euclidean distance between the trained weight and the initial (BMU) weight

7.77

7.37

Variance of the trained weightsa

0.012

0.014

  1. aVariance of the input time series was 0.017 and the variance of the initial BMU weights was 0.010.
Back to article page