Abstract: A plurality of potential feed objects and corresponding identifications of actors who performed a user interface action that caused a corresponding potential feed object to be generated are obtained. The plurality of potential feed objects and corresponding actor identifications are then fed into a machine learned feed object ranking model, with the machine learned feed object ranking model having been trained via a machine learning algorithm to calculate a score for each of the potential feed objects. The score is based on a combination of a likelihood that the user will perform an interaction, via the user interface, on the potential feed object, likelihood that the user's interaction will cause one or more downstream events by other users, and likelihood that a response from a viewer will cause the actor corresponding to the potential feed object to perform an additional user interface action to generate another potential feed object.

Abstract: Techniques for optimizing machine-learned models based on dwell time of network-transmitted content items are provided. In one technique, impression data and selection data are used train a selection prediction model. For each impression, a dwell time associated with that impression is determined and compared to a skip time. If the dwell time is less than the skip time, then a first training label that indicates that the impression is skipped is associated with the impression. If the dwell time is greater than the skip time, then a second training label that indicates that the impression is not skipped is associated with the impression. These training labels are used to train a skip prediction model. The selection prediction model and the skip prediction model are used in a content item selection event to generate a score for each candidate content item. The scores are used to select a content item.

Abstract: A plurality of potential feed objects and corresponding identifications of actors who performed a user interface action that caused a corresponding potential feed object to be generated are obtained. The plurality of potential feed objects and corresponding actor identifications are then fed into a machine learned feed object ranking model, with the machine learned feed object ranking model having been trained via a machine learning algorithm to calculate a score for each of the potential feed objects. The score is based on a combination of a likelihood that the user will perform an interaction, via the user interface, on the potential feed object, likelihood that the user's interaction will cause one or more downstream events by other users, and likelihood that a response from a viewer will cause the actor corresponding to the potential feed object to perform an additional user interface action to generate another potential feed object.