Dataset Selection #10
-
|
How did you select the different datasets? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
The selection process was application-driven. We reached out to computational scientists and numerical software developers and asked for either tasks that were relevant to their research or that they found to be challenging test cases for existing software. From there, since our goal was to develop a ML-accessible dataset, we filtered out cases where the primary challenge was scale or geometry or where there simply wasn’t enough data. To keep the focus on dynamics, we ended up selecting only systems that could be sampled reasonably on uniform space and time grids. |
Beta Was this translation helpful? Give feedback.
The selection process was application-driven. We reached out to computational scientists and numerical software developers and asked for either tasks that were relevant to their research or that they found to be challenging test cases for existing software. From there, since our goal was to develop a ML-accessible dataset, we filtered out cases where the primary challenge was scale or geometry or where there simply wasn’t enough data. To keep the focus on dynamics, we ended up selecting only systems that could be sampled reasonably on uniform space and time grids.