search

usail-hkust / t-PatchGNN

Official implementation for ICML24 paper "Irregular Multivariate Time Series Forecasting: A Transformable Patching Graph Neural Networks Approach"
37 stars 4 forks source link

Training epoch #4

Open Xyhi opened 1 month ago

Xyhi commented 1 month ago

Dear authors, I really appreciate your amazing work. May I ask if the training epochs are all set as 1000 in the experiments?

willzhang3 commented 1 month ago

All models apply the same early stopping strategy when the validation loss doesn’t decrease over 10 epochs. Therefore, the total training epochs may vary for different models and usually less than 1000.