The task is to design a feed ranking algorithm but at the step of collecting training data, the best choice seems to suggest doing it with another feed ranking algorithm.
Probably it meant if we already have an existing feed ranking algorithm to use and assumed it would in general a better choice than the other two? And if we really don’t have anything like that, we would still need to start with either one of the other two, right?
Hi, @Tyge
We are using a feed ranking algorithm for collecting training data. You are right. If we already have an existing feed ranking algorithm to use, it would be a better choice than the other two,It is recommended because it has already been optimized based on criteria like user engagement or relevance. This allows leveraging existing knowledge and insights gained during its development. However, if no existing algorithm is accessible, other options must be considered. These include ranking by chronicle order and random serving.
For model training, we’ll use different algorithms, such as Logistic regression or Neural Networks with sigmoid activation.