search

kdgutier / esrnn_torch

MIT License
156 stars 45 forks source link

how to check fitting performance #19

Closed ramdhan1989 closed 3 years ago

ramdhan1989 commented 4 years ago

is there any way to check whether the model is overfit visually ? need advice to avoid overfit model using this repo.

thank you

kdgutier commented 3 years ago

Hi @ramdhan1989, sorry for the late response. For the moment we don't have a visualization of the overfit metrics but we added evaluation during training to the model that you can plot.

As a quick advice there is a very strong regularization technique for regression problems called "early stopping" that you can track if you plot the validation loss against the gradient steps to decide where to cut the training.

Yao, Y., Rosasco, L., & Andrea, C. (2007). On early stopping in gradient descent learning. Constructive Approximation, 26(2), 289{315. URL: https://doi.org/10.1007/s00365-006-0663-2.