xiaolan98 / bdann-ijcnn2020 Goto Github PK

About the twitter dataset The question of the true and false numbers, that is, how did 7021 5974 come about and why did we reproduce a different result?

First of all, thank you very much for sharing your work on BDANN. After reading your paper, I found that the EANN and MVAE models are not well reproduced, and many multimodal fake news detection works use them as benchmarks. Could you please share the code for these two control experiments, thank you very much again.

In the above line, you use the validation directory to form the test_data.

When I changed it to the test directory, I got a very poor conformance, like 50% accuracy. Did the paper report the accuracy of validation/test sets on Twitter dataset?

Hi author! Your works are fantastic. I check your codes and data in the repo, and have a few questions as below:

• Is mediaeval 2015 or mediaeval 2016 do you use in your experiment? Or just a mixture of them? Seems that you only quote mediaeval 2015 in your paper, but I found images from 2016 test set in your data.
• You wrote loss = class_loss - domain_loss in your codes, but I think it should be loss = class_loss + domain_loss, is this normal? Why?

Hello author, in the experimental code of the paper, is the test set you placed also validate? Do you need to change it to test? Did you separate the image_validation data separately? thank you for answering me

作者您好，请问在论文的实验代码中，您放置的测试集也是validate嘛？还需要改成test嘛？image_validation数据是您单独分出来的嘛？

Hi. thank you for sharing your code. I'm trying to run BDANN_twitter.py but it faces an error line 144, in read_post
texts = open(pre_path + 'train_posts.txt', 'r').readlines()
return codecs.charmap_decode(input,self.errors,decoding_table)[0]
UnicodeDecodeError: 'charmap' codec can't decode byte 0x9d in position 6314: character maps to . could you tell me what changes is needed to get rid of this error?