Age | Commit message (Collapse) | Author | |
---|---|---|---|
2020-03-06 | Reorganize matching_dist code, minor fixes. | Arnur Nigmetov | |
2020-03-04 | Make matching distance header-only. | Arnur Nigmetov | |
2020-03-01 | Keep only unique positions in module. | Arnur Nigmetov | |
Other minor changes: 1. Remove bi-grades and unnecessary IntPoint class. 2. Remove ModulePresentation ctor from file. | |||
2020-02-19 | Require first file line specify the format. | Arnur Nigmetov | |
1. Input format must be in the first line of input file, as in RIVET. 2. Disable more unnecessary output. 3. Get rid of some warnings. | |||
2020-02-19 | Fix BifiltrationProxy bug; disable debug output. | Arnur Nigmetov | |
1. Same bifiltration was used in constructing two bifiltration proxies; the second bifiltration was ignored. Fixed. 2. Switched from spd::info to spd::debug in most places. 3. Get rid of absolute paths for test bifiltrations. | |||
2020-02-18 | Rename enum. | Arnur Nigmetov | |
2020-02-18 | Merge branch 'matching_distance' of bitbucket.org:grey_narn/hera into ↵ | Arnur Nigmetov | |
matching_distance | |||
2020-02-18 | README and boost/progress changes from master. | Arnur Nigmetov | |
2020-02-18 | Interim: matching distance for modules | Arnur Nigmetov | |
1. Templatize DistanceCalculator (DiagramProvider) 2. Add BifiltrationProxy with the same interface as ModulePresentation (dimension fixed). 3. Call Hera with relative error. 4. Add class ModulePresentation. To-Do: reading module presentations from Rivet format. | |||
2020-02-18 | Add matching distance code. | Arnur Nigmetov | |
2020-02-18 | One gitignore | Arnur Nigmetov | |
2020-02-18 | Rename directories for bottleneck and Wasserstein | Arnur Nigmetov | |
2020-01-14 | Interim: matching distance for modules | Arnur Nigmetov | |
1. Templatize DistanceCalculator (DiagramProvider) 2. Add BifiltrationProxy with the same interface as ModulePresentation (dimension fixed). 3. Call Hera with relative error. 4. Add class ModulePresentation. To-Do: reading module presentations from Rivet format. | |||
2020-01-02 | remove unused attribute | Arnur Nigmetov | |
2019-12-03 | Add matching distance code. | Arnur Nigmetov | |
2019-12-03 | One gitignore | Arnur Nigmetov | |
2019-12-03 | Rename directories for bottleneck and Wasserstein | Arnur Nigmetov | |
2019-09-17 | Add traits for long double (Wasserstein). | Arnur Nigmetov | |
2019-09-16 | Add traits for float in Wasserstein. | Arnur Nigmetov | |
2019-09-15 | Update README, BSD license only. | Arnur Nigmetov | |
2019-02-06 | Switch to opts, tolerate max-iterations-exceeded. | Arnur Nigmetov | |
1. Use opts.h for command-line parsing in wasserstein_dist. 2. Real relative error at the end of auction is stored in params, params is passed by reference. 3. If -e command line option is given to wasserstein_dist, relative error will be printed. 4. If -t option is given to wasserstein_dist, no exception will be thrown, if maximum number of iterations is exceeded. 5. Run wasserstein_dist -h to see all options. | |||
2018-06-21 | Bug in longest edge for negative persistence fixed | Arnur Nigmetov | |
2018-06-19 | Template bug fixed | Arnur Nigmetov | |
2018-06-19 | Longest edge for bottleneck, tests for bottleneck | Arnur Nigmetov | |
2018-06-03 | First tests added | Arnur Nigmetov | |
Preparation for longest edge computation. | |||
2018-06-03 | Add const qualifier | Arnur Nigmetov | |
2018-06-01 | First trivial test for point cloud version | Arnur Nigmetov | |
2018-05-31 | 2 bugs fixed in Jacobi for point cloud | Arnur Nigmetov | |
2018-05-31 | Fix bug in point cloud version. | Arnur Nigmetov | |
internal_p was not respected by DynamicPointTraits, l_2 distance was always used. | |||
2018-05-19 | Add inline qualifier to Wasserstein code. | Arnur Nigmetov | |
To avoid linking problems. | |||
2018-05-19 | Points at infinity handled correctly. | Arnur Nigmetov | |
1. Reader copied from wasserstein code, can parse inf, moved to hera namespace. 2. addProjections removes points at infinity. 3. Points at infinity handled separately by getInfinityCost(). | |||
2018-05-15 | Fixed iterator arithmetic for small diags | Arnur Nigmetov | |
2018-03-16 | Minor changes (indentation, etc) | Arnur Nigmetov | |
2018-03-16 | Test cases for the DIPHA input reader. | Gard Spreemann | |
2018-03-15 | Support for reading persistence diagrams in the DIPHA format. | Gard Spreemann | |
2018-03-06 | Points with 0 persistence ignored in the input | Arnur Nigmetov | |
2018-01-25 | Fix for inf-only diagrams | Arnur Nigmetov | |
2018-01-20 | Major rewrite, templatized version | Arnur Nigmetov | |
2017-07-19 | Include cmath instead of math.h | Arnur Nigmetov | |
2017-07-05 | Price adjustment; maxIterNum increased | Arnur Nigmetov | |
For high Wasserstein powers price adjustment (subtract minimal price) is needed. | |||
2017-04-26 | More debug output | Arnur Nigmetov | |
2017-04-26 | More debug output | Arnur Nigmetov | |
2017-04-24 | Some more debug output | Arnur Nigmetov | |
2017-04-24 | Lots of debug output for sampling heuristic | Arnur Nigmetov | |
To fix a problem on Mac OS. | |||
2017-04-24 | Verbose timing for bottleneck added | Arnur Nigmetov | |
Timing info is printed, if VERBOSE_BOTTLENECK is #defined in def_debug_bt.h. | |||
2017-04-24 | Debug output suppressed | Arnur Nigmetov | |
2017-04-24 | Empty diagram bug for Wasserstein fixed | Arnur Nigmetov | |
2017-04-14 | Bug in Wasserstein fixed: internal_p not passed to kd-tree | Arnur Nigmetov | |
2017-04-06 | Minor changes | Arnur Nigmetov | |
2017-04-06 | Heuristic for bottleneck: estimate on sampled diagram | Arnur Nigmetov | |
Sample input diagrams and use the distance between samples as initial guess in binary search. Relevant variables: bottleneck_dist.cpp:10 useSamplingHeur (if false, heuristic is never applied) bottleneck_dist.cpp:11 heurThreshold (heuristic is not applid to diagrams with fewer points) TODO: add command-line parameters instead Sampling strategy: sort points by multiplicity, set cutting value to be the multiplicity where the increase of the next one is maximal and only keep points whose multiplicity is less then the cutting value |