search

compTAG / r-tda

Other
4 stars 6 forks source link

Potential bug using GUDHI with a distance matrix for Betti 2 #13

Open dlm opened 6 years ago

dlm commented 6 years ago

See code below for details (Note, github doesn't support uploading Rmd files, so I added a txt extension).

Betti2_lost_features.Rmd.txt

To get started:

  1. see if we can make a smaller example that illustrates the issue
dlm commented 5 years ago

From Lu Xian (Lori Ziegelmeier's student)

I used a few datasets to compare the betticount data generated by library = "Dionysus" with library = c("GUDHI","PHAT"), and it seems that there is a difference:

1. Values in betticount data generated with *library = "Dionysus"* are always larger than that of *library = c("GUDHI","PHAT")*. 
2. The difference between two versions of betticount data is more significant in dimension 1.
3. In general, the difference is more obvious as eta value gets larger.
4. The max difference from the test datasets is 15, which exsits in dimension 1 (eta = 1.5 and eta = 2).

See html below for more details (Note, github doesn't support uploading Rmd files, so I added a txt extension).

TDAlibrary.html.txt