Time-Series Visualization Process for Different Datasets using Diffusion Forcing

[alanwang8]

Hello,

I currently want to use diffusion forcing for my own time-series dataset, particularly to analyze multiple time-series signals simultaneously, similar to the multivariate forecasting you mentioned in your paper.

How can I create visualizations to illustrate the forecasting/prediction process for diffusion forcing on time series data?

Please let me know if you have any questions about my use case or need further clarification.

Thanks!

[diego-marti]

Hi there! You can launch a time series experiment with the command

python -m main +name=ts_exchange dataset=ts_exchange algorithm=df_prediction experiment=exp_prediction

The time series visualizations from Fig. 6 in the paper are created automatically in the test epoch at the end of training. You can look into the function log_timeseries_plots to modify the plots. Hope this helps!