search

RUCAIBox / RecBole

A unified, comprehensive and efficient recommendation library
https://recbole.io/
MIT License
3.31k stars 599 forks source link

Training Parameters - Negative Sapling #1919

Open tripledoubleE opened 9 months ago

tripledoubleE commented 9 months ago

Hi Teams, thank you for your great RecBole library!

I'm now working on the negative sampling on RecSys and using RecBole. I have a question about one of the training parameters "train_neg_sample_args". Have the two usages in the following same effect? 1) 'train_neg_sample_args': { "uniform": 2 },

2) 'train_neg_sample_args': { "distribution": "uniform", "sample_num": 2 }

I am sure that in the second usage, we sample 2 items per user. What I'm looking for is to sample as many negative items as each user interacted with. Does the first use do this?

I'm looking forward to hearing from you if you have time. Thank you in advance!

Fotiligner commented 9 months ago

@tripledoubleE Thanks for your attention to RecBole! Actually the first way of defining train_neg_sample_args was included in the past version of recbole. Now we usually use the second way to set related parameters for train_neg_sample_args.