Live data from Hacker News

Show HN: Dreaming up dresses via WGAN

sanealytics.com

1–4 of 4 posts

Re: Show HN: Dreaming up dresses via WGAN

#3

> This gives an RMSE (don’t worry about what this means, higher is better) of 0.28. You want a lower RMSE; the E stands for "error"

It's also odd that the RMSE increases when adding a nonlinear term. The in-sample RMSE should only decrease when adding additional terms and nonlinearities (that's where overfitting comes from).

Re: Show HN: Dreaming up dresses via WGAN

#4
post #3

> This gives an RMSE (don’t worry about what this means, higher is better) of 0.28. You want a lower RMSE; the E stands for "error"

It's also odd that the RMSE increases when adding a nonlinear term. The in-sample RMSE should only decrease when adding additional terms and nonlinearities (that's where overfitting comes from).

He didn't add the log-transform, he replaced the linear relationship with the log-transform, which would cause RMSE to increase if there were a lot of high-skewing response values, which appears to be the case from the diagram.(log-transforming the response might help)