search

lindermanlab / S5

MIT License
248 stars 43 forks source link

Irregular time sampling question #13

Closed yhl48 closed 6 months ago

yhl48 commented 6 months ago

Thanks for the amazing paper, I have a technical question about the paper: if a dataset contains samples with irregular time sampling, and the irregularity is not consistent across samples, i.e. sample 1 could be sampled at different timepoints compared to sample 2, would S5 be able to model the data well?

andrewwarrington commented 6 months ago

Hi there, yes, irregular sets of irregularly sampled data are supported!

This application is demonstrated in the pendulum branch (we kept it separate because the code gets a bit bloated, and we wanted to keep the core code as simple as possible).

In the dataloader, the observation intervals are batched and returned (see here), and we discretize the SSM accordingly here.

I'll close this issue for now, but feel free to re-open it if this isn't sufficient! :)