Reproducibility

[yhl48]

For the experiment with EigenWorms dataset, is the best result reproducible with the default arguments here https://github.com/tk-rusch/LEM/blob/main/src/eigenWorms/eigenWorms_task.py?

[tk-rusch]

We presented the one best result in the paper. I only tested it on one specific GPU with one specific pytorch version (note that seeds change throughout different machines and pytorch versions). Because the dataset is so tiny it might not give exactly the same results, so I'd suggest if you want to reproduce it, to fine-tune LEM on your machine around the hyperparameters I provided. Let me know if that worked for you.

[yhl48]

Thanks, two quick follow-up questions -- how did you set the seeds for the results in the paper and is the data split the exact same as the NeuralRDE paper?