Number of Negative samples in video recommendation system


in the Video Recommendation System Design part of the “Machine Learning System Design” course, there is an assumption in the Data size section:
" * For 1 month, we collected 15 billion positive labels and 750 billion negative labels."
my question is how did we drive 750 billion negative labels from the specification?


I have the same question. In the assumption, it says that " * On average, a user watches two videos out of 100 video recommendations.". It means 15 billion (2%) out of 750 billon recommendations were viewed. Shouldn’t it be 15 billion positive labels and 735 billion negative labels?

1 Like