[thejbills]

Dear Teams

We recently identified an issue in the dataset that was brought to our attention.

Subsequent analysis confirmed that, in a limited number of cases, data from one run was unintentionally copied into a subsequent run. The affected run pairs and the extent of the overlap are summarized in the table below. The most significant issue occurs in Exp-A Run-2, which contains substantial overlap with Exp-A Run-1. The remaining instances involve smaller levels of duplication, ranging from approximately 1% to 27% of a run’s files.

We ask all teams to be aware of these overlaps when developing, training, and evaluating their models. While this issue is unfortunate, we believe the dataset continues to provide substantial value for algorithm development and benchmarking due to its size, diversity, and overall information content.

Importantly, the final evaluation will be performed on experiments that are not included in the training or test datasets. As a result, the integrity of the final scoring process remains unaffected, and we do not expect this issue to compromise the overall objectives of the challenge.

This message will also be posted on the forum. Please conduct any follow-up discussion there.

Regards,
The PHM Data Challenge committee

Folder X (source)         Folder Y (dest. error)    Number of Duplicate Files           Percentage

Exp-A/HDF5/Run-1          Exp-A/HDF5/Run-2                                311                 71.8
Exp-B/HDF5/Run-1          Exp-B/HDF5/Run-2                                154                 26.1
Exp-F/HDF5/Run-4          Exp-F/HDF5/Run-6                                120                 15.4
Exp-C/HDF5/Run-10         Exp-C/HDF5/Run-9                                104                 16.3
Exp-E/HDF5/Run-4          Exp-E/HDF5/Run-5                                 82                 12.4
Exp-C/HDF5/Run-5          Exp-C/HDF5/Run-7                                 81                 12.2
Exp-F/HDF5/Run-2          Exp-F/HDF5/Run-4                                 78                  8.7
Exp-F/HDF5/Run-1          Exp-F/HDF5/Run-2                                 55                  7.2
Exp-C/HDF5/Run-5          Exp-C/HDF5/Run-6                                 26                  3.6
Exp-F/HDF5/Run-4          Exp-F/HDF5/Run-7                                 14                  2.4
Exp-E/HDF5/Run-11         Exp-E/HDF5/Run-12                                11                  1.5
Exp-C/HDF5/Run-5          Exp-C/HDF5/Run-8                                  5                  0.7
 

For reference, the total number of TDMS files available for each experiment is:
 
Exp-A:  1,883
Exp-B:  2,375
Exp-C:  4,460
Exp-D:  2,164
Exp-E:  4,941
Exp-F:  2,916

• This topic was modified 2 hours, 33 minutes ago by thejbills.

[Author]

Posts