geoelements
Application to 3D dataset
Hello, I am interested in applying the GNS approach to my dataset. Specifically, I intend to use ceramic fragments as nodes to predict the location of the fragments after the ceramic is impacted. However, after expanding my test dataset to 3D, I am having a bit of a problem with the dimensionality of the boundary distances, and the boundaries are never dimensionally aligned with the most_recent_position. Can you suggest how best to modify the current framework to accommodate the 3D dataset? Any guidance or advice would be appreciated.
We have data format documentation here. If the data is 3D, the "bounds" keyword in metadata should have 3 items where each item corresponds to the range of boundary for each dimension ((e.g., "bounds": [[0.1, 0.9], [0.1, 0.9], [0.1, 0.9]])). Also, please check if the node feature length nnode_in aligns with your data.
We have data format documentation here. If the data is 3D, the "bounds" keyword in metadata should have 3 items where each item corresponds to the range of boundary for each dimension ((e.g.,
"bounds": [[0.1, 0.9], [0.1, 0.9], [0.1, 0.9]])). Also, please check if the node feature lengthnnode_inaligns with your data.
I created a dataset myself, where the bounds are [[0, 10], [0, 10], [0, 40]]. However, during training, I encountered all loss values being NaN. Do these bounds and positions need to be normalized before use?
