How do you get baselines_pr?

Question

How do you get baselines_pr?

WindChimeRan opened this issue 5 years ago · comments

I want to compare my model (maybe in PyTorch) with RESIDE and baselines. However, reimplementing every model is tedious. RESIDE and baselines are evaluated in GIDS, a new dataset, so I am curious where the implementations of baselines come from.

In other words, how do you get baselines_pr?

Thanks

Shikhar Vashishth · Answer 1 · Fri Dec 21 2018 18:17:24 GMT+0800 (China Standard Time)

For Riedel dataset, I got the precision-recall values for MIMLRE, Mintz and MultiR from here. For PCNN and PCNN+ATT I used the code provided by the authors. For BGWA, I used my own implementation.

Ranran · Answer 2 · Sat Dec 22 2018 08:00:56 GMT+0800 (China Standard Time)

Thanks, that's very helpful.

KuangJun · Answer 3 · Thu Feb 14 2019 10:04:21 GMT+0800 (China Standard Time)

Hi, @svjan5
Sorry to disturb you, but the baseline BGWA confuse me for a few weeks, I implemented the BGWA myself but can't get a good result. The code provided by the author is Pytorch, and I can't understand it. I try to run the source code but I'm failed. Is it possible that show me your implementation about BGWA algorithm? Thank you very much.

Shikhar Vashishth · Answer 4 · Thu Feb 14 2019 12:31:01 GMT+0800 (China Standard Time)

Hi @CrisJk,
I have included code for BGWA model here. Please have a look.

KuangJun · Answer 5 · Thu Feb 14 2019 12:33:47 GMT+0800 (China Standard Time)

Hi @CrisJk,
I have included code for BGWA model here. Please have a look.

It' very kind of you, Thank you very much.

yangluojun · Answer 6 · Tue Dec 15 2020 11:46:42 GMT+0800 (China Standard Time)

Hi, @svjan5
Sorry to disturb you, but the baseline BGWA confuse me for a few weeks, I implemented the BGWA myself but can't get a good result. The code provided by the author is Pytorch, and I can't understand it. I try to run the source code but I'm failed. Is it possible that show me your implementation about BGWA algorithm? Thank you very much.

I also got a worser result by running the BGWA code provided the repository than the orgial author, but if I run the code provided by the original author, the result will be good. Have you made it clear? Thanks.

Shikhar Vashishth · Answer 7 · Tue Dec 15 2020 22:43:10 GMT+0800 (China Standard Time)

Hi, I did not thoroughly check the baseline code. For the numbers in the paper, I used the original authors' code. The BGWA code which is part of this repository was written only for the sake of helping the researchers. For PCNN and PCNN+ATT baseline, I used the original c++ based code of the authors. Please feel free to open pull request if you find any error in the code which should be corrected.

Thanks!