search

declare-lab / SANCL

[COLING 2022] This repository contains the code of the paper SANCL: Multimodal Review Helpfulness Prediction with Selective Attention and Natural Contrastive Learning.
MIT License
7 stars 0 forks source link

Dissertation reproduction work questions #3

Closed APPLE-XMT closed 1 year ago

APPLE-XMT commented 1 year ago

Hello, it is a great honor to read your article. When I am doing the recurrence work, I always use the learning rate in the config file to achieve the best results in the 1-2 epoch. Is this correct?

Clement25 commented 1 year ago

It's very normal to converge within 5 epochs since the dataset is very huge.

APPLE-XMT commented 1 year ago

ok thanks for your answer

APPLE-XMT commented 1 year ago

Hello, I see that you mentioned in the paper that the label range of the dataset is 0-4, but I found that there is no label of 0 in the amazon dataset. Why is this?

Clement25 commented 1 year ago

I also found this and don't know either.