Why can multiple output networks help when strong correlation exist

Q: Assuming there is a strong correlation between how long users spend on listings and the likelihood of booking, how would you redesign the network architecture?

A: Train multiple-output networks for two outputs: view_time and booking


Hi @Geoffrey_Hung,

We’re talking about the time that users spend on listings and booking for the Possibility of booking. That is why we will train the multiple output networks for view_time and booking.