Recommendation Systems / Engines with TensorFlow - Google Cloud Platform User Group Singapore

[@ninatsvetkova]

Very helpful! I also recommend the paper MATRIX FACTORIZATION TECHNIQUES FOR RECOMMENDER SYSTEMS by Yehuda Koren, Robert Bell and Chris Volinsky for people (like me) that are very new in the field of recommendation systems.

[@geoffreyanderson4719]

That's a great paper to start with. And for a concise academic video of the central model described in that paper, please consider to watch the Module 5 videos of Mining Massive Datasets course by profs at Stanford. The site is lagunita dot stanford dot edu. I recently used these resources as basis for a good tensorflow implementation.

[@geoffreyanderson4719]

How does the lecturer use SVD to factorize the ratings data, when the ratings data is mostly missing (nobody watches all the movies), and SVD does not work on missing data at all? Are embeddings the key?

[@geoffreyanderson4719]

Splitting into 90% train 10% test leaves NO way to find an unbiased estimate of the final model's error (37m57s). It's a basic mistake. Should have held some examples aside for error estimation on the final model.

[@blank8026]

Poor explaination!