search

ts-kim / RevIN

RevIN: Reversible Instance Normalization For Accurate Time-series Forecasting Against Distribution Shift
MIT License
272 stars 23 forks source link

Hi, when will the code be updated? #1

Closed sou6qiu closed 2 years ago

sou6qiu commented 2 years ago

I have difficulty implementing RevIN on Informer. I'm confused about what to do with the Decoder input of Informer, so I would like to refer to the official code.

ts-kim commented 2 years ago

Thank you for your interest in our work. We have uploaded the official code.

The training and evaluation codes for Informer will be made available shortly.

sou6qiu commented 2 years ago

Hi again. I appreciate you taking the time to respond to my question. I've been able to run RevIN on Informer thanks to your code, but I still have some questions about how RevIN works. Did you use RevIN after standardization or just remove the standardization in Informer? [image: image.png] I tried both two methods and the latter turned out to be much less effective than the former. So I wonder how you deal with that.

ts-kim @.***> 于2022年5月17日周二 22:34写道：

Thank you for your interest in our work. We have uploaded the official code.

You can use RevIN as following:

from RevIN import RevIN revin_layer = RevIN(num_features)

x_in = revin_layer(x_in, 'norm') x_out = informer(x_in) # your model x_out = revin_layer(x_out, 'denorm')

Hope it works out. :)

— Reply to this email directly, view it on GitHub https://github.com/ts-kim/RevIN/issues/1#issuecomment-1128949521, or unsubscribe https://github.com/notifications/unsubscribe-auth/AK5SRBRPXRNWZTAN5LWRRDDVKOU6PANCNFSM5V7CJP4A . You are receiving this because you authored the thread.Message ID: @.***>

zhang-potato commented 2 years ago

I have difficulty implementing RevIN on Informer. I'm confused about what to do with the Decoder input of Informer, so I would like to refer to the official code.

Hello, I would like to ask how you added the RevIN in the Informer. The effect seems to be not obvious after I added it. Can you show me how you added the RevIN? My email is 15993617315@163.com, thank you very much.

seharanul17 commented 2 years ago

Hi again. I appreciate you taking the time to respond to my question. I've been able to run RevIN on Informer thanks to your code, but I still have some questions about how RevIN works. Did you use RevIN after standardization or just remove the standardization in Informer? [image: image.png] I tried both two methods and the latter turned out to be much less effective than the former. So I wonder how you deal with that.

We used RevIN after z-score standardization in Informer. Since instance normalization is employed when RevIN is performed at the input layer, the normalized samples in a batch stay the same regardless of the presence of standardization. However, performance differences can exist because the value of gradients can still be influenced.

seharanul17 commented 2 years ago

Hello, I would like to ask how you added the RevIN in the Informer. The effect seems to be not obvious after I added it. Can you show me how you added the RevIN? My email is 15993617315@163.com, thank you very much.

Hello, thank you for your interest in our work. We have uploaded the RevIN codes for Informer. Please check it.