RippleNet using Wikidata and Movielens #1014
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@almudenasanz, this is the error I got: |
@miguelgfierro |
|
@almudenasanz I fixed the error, now everything computes correctly, however, the metrics looks too high, we should take a look |
|
so it looks that the problem is that we are scoring items that are already seen, that's why p@k and r@k are so high. In other algos, we have what we call Almudena, can you please take a look at the remove seen example? |
|
hey @almudenasanz, @loomlike found out that in the function but in movielens, there are no zeros, it goes from 1 to 5. So there has to be some error there |
|
In introduced new parameters for two functionalities:
I think we can run it with 1M and it would be ready |
|
hey @almudenasanz I rerun with 100k movielens, still there is an error in the ranking metrics. Please take a look |
|
hey @almudenasanz, this PR has been active since December. I think we should split this PR to have the AUC and accuracy that matches the paper and merge it. Later, on a different PR, we can take a look at the ranking metrics that are not providing good results. |
|
Hi @miguelgfierro I agree, I would just remove the cells with the precision_at_k and recall_at_k and reproduce with 1M |
|
Hey @miguelgfierro! Do we plan to release the Ripple-Net without the Thanks Edits -> I wish the src was in PyTorch rather than tf 🤐 . Secondly, How to get access to Wiki's KG? |
|
hey @AdityaSoni19031997, answering your questions:
that sounds good
This notebook was done years ago when we were using TF1, now we are using TF2. You could either use TF2 or if you want, implement it in PyTorch (there is an implementation https://github.com/qibinc/RippleNet-PyTorch). In general, we prefer PyTorch over TF.
The KG is generated from this notebook https://github.com/microsoft/recommenders/blob/ad0bc93142586dd1b856f9bb8629c89947c93235/notebooks/01_prepare_data/wikidata_knowledge_graph.ipynb |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
Notebook showcasing RippleNet https://github.com/hwwang55/RippleNet and related files
Uses Movielens dataset and Wikidata as the KG
Reproduces results of https://arxiv.org/abs/1803.03467 for Movielens 1M using Wikidata instead of Microsoft Satori as the KG
Need to update tests
Related Issues
Checklist: