Abstract: Techniques for modifying training data for video response quality optimization are provided. In one technique, training data is identified that is generated based on video presentation data that indicates multiple video items were presented to multiple entities. The training data comprises multiple training instances, each indicating a presentation of at least a portion of a video item to an entity. For each training instance in a subset of the training instances, a quality metric of the presentation of the video item indicated in said each training instance is computed and that training instance is modified based on the quality metric. After modifying one or more of the training instances, the model is trained using one or more machine learning techniques. In response to a content request, the model is used to determine whether to transmit a particular video item over a network to a computing device of a particular entity.

Abstract: Techniques for modifying training data for video response quality optimization are provided. In one technique, training data is identified that is generated based on video presentation data that indicates multiple video items were presented to multiple entities. The training data comprises multiple training instances, each indicating a presentation of at least a portion of a video item to an entity. For each training instance in a subset of the training instances, a quality metric of the presentation of the video item indicated in said each training instance is computed and that training instance is modified based on the quality metric. After modifying one or more of the training instances, the model is trained using one or more machine learning techniques. In response to a content request, the model is used to determine whether to transmit a particular video item over a network to a computing device of a particular entity.

Abstract: Systems and methods for optimizing offsite content delivery are provided. A content request is received from a content exchange and multiple candidate content delivery campaigns are identified in response to the content request. A computerized method includes, for each candidate content delivery campaign, determining a resource usage per conversion on a particular content platform, determining a conversion rate on one or more third-party content platforms, and determining a conversion rate on the one or more third-party content platforms. The resource usage per impression is computed based on the resource usage per conversion, the resource usage per selection, and the conversation rate. A particular candidate content delivery campaign is selected from among multiple candidate content delivery campaigns based on the resource usage per impression and the particular candidate content delivery campaign is caused to be transmitted over a computer network to the content exchange.

Abstract: Techniques for predicting an offsite entity interaction rate are provided. One approach involves using a first machine-learned model that includes a first plurality of features that correspond to entity and campaign attributes. The approach also involves training a second machine-learned model that includes a second plurality of features that includes a particular feature corresponding to predicted entity interaction rates. Thus, output of the first machine-learned model is input to the second machine-learned model. The second machine-learned model includes multiple weights that include a particular weight for the particular feature. A content request is received and a set of campaigns is identified based on an entity identifier associated with the content request. Scores are generated based on the first and second machine-learned models. Based on the scores, a campaign is selected and campaign data associated with the campaign is transmitted over a computer network.

Abstract: Systems and methods for optimizing offsite content delivery are provided. A content request is received from a content exchange and multiple candidate content delivery campaigns are identified in response to the content request. A computerized method includes, for each candidate content delivery campaign, determining a resource usage per conversion on a particular content platform, determining a conversion rate on one or more third-party content platforms, and determining a conversion rate on the one or more third-party content platforms. The resource usage per impression is computed based on the resource usage per conversion, the resource usage per selection, and the conversation rate. A particular candidate content delivery campaign is selected from among multiple candidate content delivery campaigns based on the resource usage per impression and the particular candidate content delivery campaign is caused to be transmitted over a computer network to the content exchange.

Abstract: Techniques for predicting an offsite entity interaction rate are provided. One approach involves using a first machine-learned model that includes a first plurality of features that correspond to entity and campaign attributes. The approach also involves training a second machine-learned model that includes a second plurality of features that includes a particular feature corresponding to predicted entity interaction rates. Thus, output of the first machine-learned model is input to the second machine-learned model. The second machine-learned model includes multiple weights that include a particular weight for the particular feature. A content request is received and a set of campaigns is identified based on an entity identifier associated with the content request. Scores are generated based on the first and second machine-learned models. Based on the scores, a campaign is selected and campaign data associated with the campaign is transmitted over a computer network.

Abstract: Machine-learned models are selectively distributed to a plurality of computer servers according to conditions associated with the computer servers. A server receives travel information from a travel coordination system. The travel information describes a plurality of conditions. The server identifies a hierarchy of one or more parent-child relationships based on the plurality of conditions. The server trains machine-learned models using the plurality of conditions described by the travel information. The server selects machine-learned models for the plurality of conditions responsive to the identified hierarchy. The server distributes machine-learned models to the plurality of computer servers responsive to the identified hierarchy.

Abstract: Methods and apparatus for identifying on-line users for advertisement or content targeting are disclosed. Historical user data is obtained in association with user identifiers, which have been unambiguously determined. The historical user data includes event data for one or more on-line user events that have occurred for each user identifier. The historical user data also specify fingerprint vectors of characteristic values that are each associated with specific ones of the user identifiers. A current one of the fingerprint vectors that is ambiguously associated with two or more user identifiers is received. A first user identifier is selected from the associated two or more user identifiers of the current fingerprint vector based on the event data of the historical user data. The selected first user identifier is provided to a server configured to provide advertisement or content based on user profile data that is obtainable for such selected first user identifier.

Abstract: Machine-learned models are selectively distributed to a plurality of computer servers according to conditions associated with the computer servers. A server receives travel information from a travel coordination system. The travel information describes a plurality of conditions. The server identifies a hierarchy of one or more parent-child relationships based on the plurality of conditions. The server trains machine-learned models using the plurality of conditions described by the travel information. The server selects machine-learned models for the plurality of conditions responsive to the identified hierarchy. The server distributes machine-learned models to the plurality of computer servers responsive to the identified hierarchy.

Abstract: Methods and apparatus for identifying on-line users for advertisement or content targeting are disclosed. Historical user data is obtained in association with user identifiers, which have been unambiguously determined. The historical user data includes event data for one or more on-line user events that have occurred for each user identifier. The historical user data also specify fingerprint vectors of characteristic values that are each associated with specific ones of the user identifiers. A current one of the fingerprint vectors that is ambiguously associated with two or more user identifiers is received. A first user identifier is selected from the associated two or more user identifiers of the current fingerprint vector based on the event data of the historical user data. The selected first user identifier is provided to a server configured to provide advertisement or content based on user profile data that is obtainable for such selected first user identifier.