Fig. 1
From: BayesProject: Fast computation of a projection direction for multivariate changepoint detection
Dependence of BayesProject (black crosses), Inspect (red circles), sum-cusum (green squares) and max-cusum (blue triangles) on the number of series p while \(n=10{,}000\). Rows show proportion of estimated changes (first row), average distance to the true changepoint location (second row), \(L_2\) distance between the estimated and ideal projection directions (third row) and runtime (fourth row). Columns show sparse (left), moderate (middle) and dense (right) scenarios. Log–log plots to assess runtime scalings in the fourth row. Slope estimates for the moderate scenario (middle column) are 1.23 (Inspect), 1.00 (BayesProject), 1.03 (sum-cusum) and 1.03 (max-cusum). (Color figure online)