How to Improve Similarity in Your Recommendations

You can get diversity from a lot of different ways so i guess we're talking just about embeddings and like these types of encoder only models. You can achieve this by for example changing up dropout a little bit you can sort of like maybe provide some sort of like noise context to these types of models which i don't think that's something people do but it's entirely possible if you want to take the idea from gans. Trying to build better content based embeddings is a way if say you don't have amazon's voluminous data in order to try and level up the playing fieldthat would be one step. Vodie: I wouldn't worry too