new dataset

[JIAYI-world]

Thank you very much for your timely help. The problem of the previous dataset has been solved, and now I have encountered two problems again. I really hope to get your help.
First of all, I ran it on my computer, and found that there were certain errors in the results of amazon data sets and kwai data sets compared with those in the paper. Mainly in the hot project and the overall project, the results were lower than those in the paper.
Secondly, I would like to run on new data sets, could you provide some data set processing methods?
Thank you so much.

[Linxyhaha]

Hi,
For your first question, the different results may be due to the pytorch version and GPU. Please make sure the pytorch version is correctly selected in your environment. We ran our experiments with pytorch 1.10.0 on A5000.
For the second question, please refer to #1 (comment) for instructions on using new datasets.

[JIAYI-world]

I apologize for not responding earlier as I only just opened the project today due to some reasons. I am glad to receive your reply.
However, I have some confusion regarding the img_pca_map in the dataset processing. Could you please explain how the 64-dimensional feature vector was obtained?
I look forward to your reply.

[Linxyhaha]

I apologize for not responding earlier as I only just opened the project today due to some reasons. I am glad to receive your reply. However, I have some confusion regarding the img_pca_map in the dataset processing. Could you please explain how the 64-dimensional feature vector was obtained? I look forward to your reply.

Hi, we apply PCA to the original visual feature to obtain the 64-dimensional feature vector.

[JIAYI-world]

Do features are extracted directly from images, but my current dataset does not have pictures or videos,only contains some user clicks, video playback time, video tags, etc., how to extract feature vectors in this case?