Hyper-parameters Suggestion for Other Datasets

wuliwei9278 / SSE-PT

Codes and Datasets for paper RecSys'20 "SSE-PT: Sequential Recommendation Via Personalized Transformer" and NurIPS'19 "Stochastic Shared Embeddings: Data-driven Regularization of Embedding Layers"

106 stars 18 forks source link

Hyper-parameters Suggestion for Other Datasets #1

Open AaronHeee opened 3 years ago

AaronHeee commented 3 years ago

Thanks for this nice work!

I'm wondering if we could get the hyper-parameter settings for other datasets (e.g. Beauty, Steam), and if there are some suggestions for hyper-parameters tuning (e.g. item/user_threshold)?

Looking forward to your reply and thanks in advance!

wuliwei9278 commented 3 years ago

I wrote down the hyper-parameters back then in 2019 but I have not touched the codes for a while now... and my Google cloud instances got shut down already, so I am not sure if I can find them right now.... but I don't think it is that hard to tune parameters to get the results in the paper. sorry!

ahmedrashed-ml commented 3 years ago

Dear Authors,

I am also facing the same issue on the Games dataset. I tried different configurations but the best performance achievable is HR@10 = 0.748 , NDCG@10 = 0.545 which is the same as SASRec. Is it possible to share with us the best parameter you used on this dataset in order to replicate your results?