diff options
| author | Arnur Nigmetov <a.nigmetov@gmail.com> | 2017-04-04 14:19:55 +0200 |
|---|---|---|
| committer | Arnur Nigmetov <a.nigmetov@gmail.com> | 2017-04-04 14:19:55 +0200 |
| commit | 2b48644a5a7d0549cc1d8e5f5cd05da984bc0669 (patch) | |
| tree | b45ec209512643c42c4683e5b3a51a92d0ebae4a | |
| parent | 404048e9cca7fa77d2004875ffdae1eab39d49f4 (diff) | |
README edited: relative_error clarified
| -rw-r--r-- | geom_bottleneck/README | 9 | ||||
| -rw-r--r-- | geom_matching/README | 7 |
2 files changed, 15 insertions, 1 deletions
diff --git a/geom_bottleneck/README b/geom_bottleneck/README index 8dda148..839ffa5 100644 --- a/geom_bottleneck/README +++ b/geom_bottleneck/README @@ -17,7 +17,14 @@ Your compiler must support C++11. bottleneck_dist file1 file2 [relative_error]. -If relative error is not supplied, the exact distance is computed. +If relative error is not supplied, the exact distance is computed and printed. +If two diagrams are equal, then the exact distance 0.0 is printed (the order +of points in file1 and file2 need not be the same). +Otherwise the output is an approximation of the exact distance. Precisely: +if d_exact is the true distance and d_approx is the output, then + + | d_exact - d_approx | / d_exact < relative_error. + file1 and file2 must contain persistence diagrams in plain text format (one point per line, empty lines are ignored, comments can be made with #): diff --git a/geom_matching/README b/geom_matching/README index 33a5611..fd987a1 100644 --- a/geom_matching/README +++ b/geom_matching/README @@ -24,6 +24,13 @@ wasserstein_dist file1 file2 [wasserstein degree] [relative error] [internal no Parameter wasserstein degree corresponds to $q$, when it tends to infinity, Wasserstein distance tends to the bottleneck distance. +If two diagrams are equal, then the exact distance 0.0 is printed (the order +of points in file1 and file2 need not be the same). +Otherwise the output is an approximation of the exact distance. Precisely: +if d_exact is the true distance and d_approx is the output, then + + | d_exact - d_approx | / d_exact < relative_error. + Parameter internal_p corresponds to p. Default values: |