How is ml20_pca128.pkl embedding generated and how to generate the similar embeddings for ml_1m and ml_100k datasets ?

awarebayes / RecNN

Reinforced Recommendation toolkit built around pytorch 1.7

Apache License 2.0

577 stars 114 forks source link

How is ml20_pca128.pkl embedding generated and how to generate the similar embeddings for ml_1m and ml_100k datasets ? #12

Closed honolulu2 closed 4 years ago

awarebayes commented 4 years ago

Hi, I am sorry for it taking so long to write back.

All of the embedding generation pipeline can be found there: https://github.com/awarebayes/RecNN/tree/master/examples/0.%20Embeddings%20Generation

Try using the 20M version with the lesser datasets, they probably will work regardless

awarebayes commented 4 years ago

Hello, has it worked out for you? I do not like to keep issues opened...

ucasiggcas commented 4 years ago

dear， could you pls try the Youtube2016 or Youtube2019?

awarebayes commented 4 years ago

dear， could you pls try the Youtube2016 or Youtube2019?

No, not right now at least. Kind of busy to work on it, although you contribution will be welcome. But I had already implemented the REINFORCE algorithm that was developed by google and tested on the youtube dataset.