Abstract: An online system matches a user across multiple online systems based on image data for the user (e.g., profile photo) regardless whether the image data is from the online system, a different but related online system or a third party system. For example, to match the user across a social networking system and INSTAGRAM™ system, the online system compares the similarity between images of the user from both systems in addition to similarity of textual information in the user profiles on both systems. The similarity of image data and the similarity of textual information associated with the user are used by the online system as indicators of matched user accounts belonging to the same user across both systems. The online system applies models trained using deep learning techniques to match a user across multiple online systems based on the image data and textual information associated with the user.

Abstract: A method for predicting demographic information for a common user that is associated with a plurality of unresolved identifiers. An unresolved identifier defines a context in which a client device accesses one or more online systems, the context not determined to be associated with a specific user. The method comprises identifying a set of unresolved identifiers, and identifying information describing one or more access events associated with each unresolved identifier. For each pair of unresolved identifiers, a similarity score for the pair is determined based on the identified information. Responsive to the similarity score exceeding a threshold similarity score, the pair of unresolved identifiers is clustered, the clustering indicating a prediction that the pair of unresolved identifiers are associated with a common user. Predicted demographic information is determined for each unresolved identifier in the cluster.

Abstract: Information describing deliveries of content items and user actions associated with the content items is stored. Each delivery is performed by an online publisher to a user. A user action associated with a content item performed by a target user is detected. Information describing a set of online publishers that delivered the content item to the target user is retrieved. For each online publisher of the set, a likelihood that the user action would have occurred without the online publisher's delivery of the content item to the target user is determined. An estimated increase in the likelihood that the user action occurred due to the online publisher's delivery of the content item to the target user is determined. A performance metric is determined for the online publisher, wherein ratios of performance metrics for the set of online publishers are related based on corresponding ratios of the estimated increases in likelihoods.

Abstract: An online system promotes physical store visits by presenting users with content items for a physical store location and subsequently logs visits of online system users to the physical store location to track performance of a campaign associated with the presented content item. The online system registers attention events associated with the presented content items presented to users on third party publishing sites via tracking pixels and registers attention events as store front visit conversion events if, within a predetermined period of time from a valid attention event, a user has subsequently gone in and visited the physical store front location.

Abstract: An online system determines the likelihood of an interaction between a user and a content item being an invalid interaction. The online system receives an indication of an interaction of a client device with a content item. The online system identifies a device ID for the client device and determines whether the device ID is associated with one or more browser IDs. If the device ID is not associated with any browser ID, the received interaction is likely an invalid interaction. The online system may further determine the likelihood of an online publisher manufacturing interactions. The online system determines a number of invalid interactions and a number of valid interactions associated with the online publisher. The online system determines a ratio between the number of invalid and valid interactions. If the ratio is larger than a threshold value, the online system determines that the online publisher is likely manufacturing interactions.

Abstract: A method for providing content items to one or more client devices associated with at least one unresolved identifier. An unresolved identifier defines a context in which a client device accesses one or more online systems, the context not determined to be associated with a specific user. The method comprises identifying a set of unresolved identifiers, and identifying information describing one or more access events associated with each unresolved identifier. For each pair of unresolved identifiers, a similarity score for the pair is determined based on the identified information. Responsive to the similarity score exceeding a threshold similarity score, the pair of unresolved identifiers is clustered, the clustering indicating a prediction that the pair of unresolved identifiers are associated with a common user. Based on this clustering, a content item is displayed on or more user devices associated with at least one unresolved identifier of the set of unresolved identifiers.

Abstract: Disclosed is an online system that infers interests of unresolved users for whom the interests are not known. The online system determines certain features about the unresolved users, but does not have certain information about the users themselves (e.g., their interests), so instead infers these attributes based on the features of the user. The online system provides the features as input to a classifier trained to predict a particular interest, and the classifier outputs a prediction of whether the user has the corresponding interest. In one embodiment, the online system trains a classifier for various interest values by forming training sets for the interests using the features for users who are logged into the online system and hence have known interests.

Abstract: A method for determining reach of a content item that is displayed on one or more client devices associated with at least one unresolved identifier. An unresolved identifier defines a context in which a client device accesses one or more online systems, the context not determined to be associated with a specific user. The method comprises identifying a set of unresolved identifiers, and identifying information describing one or more access events associated with each unresolved identifier. For each pair of unresolved identifiers, a similarity score for the pair is determined based on the identified information. Responsive to the similarity score exceeding a threshold similarity score, the pair of unresolved identifiers is clustered, the clustering indicating a prediction that the pair of unresolved identifiers are associated with a common user. Finally, for the reach of the displayed content item is determined based on the clustering of the set of unresolved identifiers.

Abstract: An online system matches a user across multiple online systems based on image data for the user (e.g., profile photo) regardless whether the image data is from the online system, a different but related online system or a third party system. For example, to match the user across a social networking system and INSTAGRAM™ system, the online system compares the similarity between images of the user from both systems in addition to similarity of textual information in the user profiles on both systems. The similarity of image data and the similarity of textual information associated with the user are used by the online system as indicators of matched user accounts belonging to the same user across both systems. The online system applies models trained using deep learning techniques to match a user across multiple online systems based on the image data and textual information associated with the user.

Abstract: An online system determines the likelihood of an interaction between a user and a content item being an invalid interaction. The online system receives an indication of an interaction of a client device with a content item. The online system identifies a device ID for the client device and determines whether the device ID is associated with one or more browser IDs. If the device ID is not associated with any browser ID, the received interaction is likely an invalid interaction. The online system may further determines the likelihood of an online publisher manufacturing interactions. The online system determines a number of invalid interactions and a number of valid interactions associated with the online publisher. The online system determines a ratio between the number of invalid and valid interactions. If the ratio is larger than a threshold value, the online system determines that the online publisher is likely manufacturing interactions.

Abstract: An online system matches a user across multiple online systems based on image data for the user (e.g., profile photo) regardless whether the image data is from the online system, a different but related online system or a third party system. For example, to match the user across a social networking system and INSTAGRAM™ system, the online system compares the similarity between images of the user from both systems in addition to similarity of textual information in the user profiles on both systems. The similarity of image data and the similarity of textual information associated with the user are used by the online system as indicators of matched user accounts belonging to the same user across both systems. The online system applies models trained using deep learning techniques to match a user across multiple online systems based on the image data and textual information associated with the user.

Abstract: An online system maintains characteristics for its users and may access characteristics of users maintained by a third party system. The online system may select content for a user of the third party system based on characteristics maintained by the third party system. If the third party system does not maintain a characteristic for its users, the generates a model predicting the characteristic for third party system users based on a set of online system users identified based on characteristics of third party system users. The online system clusters third party system users based on the predicted characteristic for other third party system users connected to the third party system user. Using verified characteristics for third party system users from a trusted third party system, the online system determines an accuracy of the predicted characteristic for third party system users in a cluster.

Abstract: Information describing deliveries of content items and user actions associated with the content items is stored. Each delivery is performed by an online publisher to a user. A user action associated with a content item performed by a target user is detected. Information describing a set of online publishers that delivered the content item to the target user is retrieved. For each online publisher of the set, a likelihood that the user action would have occurred without the online publisher's delivery of the content item to the target user is determined. An estimated increase in the likelihood that the user action occurred due to the online publisher's delivery of the content item to the target user is determined. A performance metric is determined for the online publisher, wherein ratios of performance metrics for the set of online publishers are related based on corresponding ratios of the estimated increases in likelihoods.

Abstract: An online system matches a user across multiple online systems based on image data for the user (e.g., profile photo) regardless whether the image data is from the online system, a different but related online system or a third party system. For example, to match the user across a social networking system and INSTAGRAM™ system, the online system compares the similarity between images of the user from both systems in addition to similarity of textual information in the user profiles on both systems. The similarity of image data and the similarity of textual information associated with the user are used by the online system as indicators of matched user accounts belonging to the same user across both systems. The online system applies models trained using deep learning techniques to match a user across multiple online systems based on the image data and textual information associated with the user.

Abstract: An online system predicts gender, age, interests, or other demographic information of a user based on image data of the user, e.g., profile photos, photos the user posts of him/herself within an online system, and photos of the user posted by other users socially connected with the user, and textual data in the user's profile that suggests age or gender (e.g., like or dislikes similar to a population of users of an online system). The online system similarly predicts a user's interests based on the photos of the user. The online system applies one or more models trained using deep learning techniques to generate the predictions. The online system uses the predictions to build more information about the user in the online system, and provide improved and targeted content delivery to the user that may have disparate information scattered throughout different online systems.

Abstract: An online system matches a user across multiple online systems based on image data for the user (e.g., profile photo) regardless whether the image data is from the online system, a different but related online system or a third party system. For example, to match the user across a social networking system and INSTAGRAM™ system, the online system compares the similarity between images of the user from both systems in addition to similarity of textual information in the user profiles on both systems. The similarity of image data and the similarity of textual information associated with the user are used by the online system as indicators of matched user accounts belonging to the same user across both systems. The online system applies models trained using deep learning techniques to match a user across multiple online systems based on the image data and textual information associated with the user.

Abstract: An online system predicts gender, age, interests, or other demographic information of a user based on image data of the user, e.g., profile photos, photos the user posts of him/herself within an online system, and photos of the user posted by other users socially connected with the user, and textual data in the user's profile that suggests age or gender (e.g., like or dislikes similar to a population of users of an online system). The online system similarly predicts a user's interests based on the photos of the user. The online system applies one or more models trained using deep learning techniques to generate the predictions. The online system uses the predictions to build more information about the user in the online system, and provide improved and targeted content delivery to the user that may have disparate information scattered throughout different online systems.

Abstract: An online system measures performance of content presented to a plurality of identifiable and non-identifiable individuals based on matching user identifying information included in data describing presentation of the content and data describing performance of an action associated with the content. To reduce measurement inaccuracy resulting from incomplete matching of user identifying information associated with non-identifiable individuals, the online system generates models to extrapolate data describing an amount of unique individuals presented with the content, an amount of unique individuals who performed an action associated with the content, and an amount of unique individuals who performed the action associated with the content attributable to presentation of the content by a content publisher. The models are applied to data collected by the online system describing presentation of the content and performance of actions associated with the content.

Abstract: Disclosed is an online system that infers demographic attributes of unresolved users for whom the demographic attributes are not known. The online system determines certain features about devices used by the unresolved users, but does not have certain information about the users themselves (e.g., their age, gender, or location), so instead infers these attributes based on the features of the user devices. The online system provides the features about the devices as input to a classifier trained to predict a particular demographic attribute value, and the classifier outputs a prediction of whether the user of the user device has the corresponding value of the demographic attribute. In one embodiment, the online system trains a classifier for various demographic attribute values by forming training sets for the demographic attribute values using the features of devices for users who are logged into the online system and hence have known demographic attribute values.

Abstract: Disclosed is a content sharing system that infers demographic attributes of users of the content sharing system based on features of the users with accounts matched to an online system with known demographic attributes. The features include attributes of unidirectional connections of the users on the content sharing system. In some embodiments, the features are distributions of demographic attributes of the unidirectional connections of the users, such as distributions of ages or genders of the unidirectional connections. The content sharing system provides the features as input to a classifier trained to predict a particular demographic attribute value and the classifier outputs a predicted value of that demographic attribute. In some embodiments, the content sharing system trains a classifier for various demographic attributes by forming training sets for the demographic attributes using the features for users.