Facebook, Inc.

Aleksandar Ilic

Florin Ratiu

David S Park

Yiyu Li

Andrew Alexander Birchall

Vasanth Kumar Rajendran

Nathan Paul Schloss

Patrick Jonathan Varin

Branislav Stojkovic

Peter Henry Martinazzi

Matthew William Kelly

John Torres Fremlin

Ziqing Mao

Chuang Wu

Ming Hua

Hong Yan

Martin Rehwald

Ariel Benjamin Evnine

Igor Kabiljo

Pradeep Kumar Sharma

Maja Kabiljo

Oleksandr Kuvshynov

Ashwin Murthy

Konstantine Oleksiyovich Kolomoyskyy

In one embodiment, a method includes receiving a number of notifications of activity relevant to a user. Each notification has an associated type. The method also includes calculating an interest of each activity to the user based at least in part on the type of each notification; ranking the notifications based at least in part on the calculated interest; and sending one or more of the notifications to the user. Each of the sent notifications has a ranking higher than a pre-determined threshold ranking.

Notification targeting

In one embodiment, a method includes sending, through a communications network, several volumes of notifications corresponding to a first notification type to multiple users and several volumes of notifications corresponding to a second notification type to multiple users. The method further determines visitation impacts of the volumes of notifications of the first and second notification types and trains a machine-learning model based on the visitation impacts. The machine-learning model generates an assessment of a likelihood of interaction by a recipient user with each of the notifications.

Value model for sending notifications

In one embodiment, a method includes receiving a number of notifications of one or more activities relevant to a user. Each notification has an associated receipt time and type of notification. The method also includes aggregating one or more of the notifications based on the type of notification; determining a sending time to send the aggregated notifications based at least in part on determining that a pre-determined amount of time that has elapsed from a receipt time of a most recent one of the aggregated notifications; and sending the aggregated notifications to the user based on the sending time.

Batching of notifications

A social networking system selects a set of contacts for presentation to a user of the social networking system. A contact is an additional user or a group of users maintained by the social networking system. To select the set of contacts, the social networking system generates scores associated with additional users of the social networking system and groups of users based on interactions between the user and various additional users and between the user and groups of users. Based on the scores associated with additional users and with groups of users, the social networking system selects a set of contacts from the additional users and the groups of users. Information identifying the set of contacts is communicated from the social networking system to a client device for presentation to the user.

Presenting contacts to a social networking system user based on user interactions with additional users and with groups

In one embodiment, a method includes one or more computing devices accessing a notification to be sent to a user, where the notification has a context. The method also includes one or more computing devices sending a request to a history service for historical notification data associated with the user with respect to the context of the notification and a ranking of the notification where the ranking indicates a probability of the user interacting with the notification. The method also includes one or more computing devices receiving the historical notification data and the ranking from the history service. Moreover, the method also includes one or more computing devices determining a notification policy to apply to the notification based at least in part on the context of the notification, the historical notification data, and the ranking. Furthermore, the method also includes one or more computing devices applying the notification policy to the notification to be sent to the user.

User-aware notification delivery

In one embodiment, a method includes detecting a triggering event for sending a notification to a user of an online social network. The method may also include accessing multiple of versions of the notification. Each version may include one or more template elements of a particular content type. The method may also access user data associated with the user that includes data associated with one or more relationships of the user on the online social network or data associated with actions performed by the user. The method may also include determining a score for each version representing a likelihood of the corresponding version being consumed by the user. The method may also include generating a personalized notification by using the selected version of the notification and replacing each of the template elements of the selected version of the notification with content of the content type associated with the template element.

Sending personalized notifications over a communication network

Embodiments are disclosed for data computation of collaborative filtering in a social network. Collaborative filtering involves predicting a user's behavior or interests based on other users' behavior or interests. To predict a user's interests in an item such as a picture, a system performs an iterative computation to perform an evaluation by solving an objective function. The system characterizes 'users' as 'vertices' in a directed graph, &#x201c;relationship among users&#x201d; as &#x201c;edges&#x201d; in the directed graph, and &#x201c;items&#x201d; as &#x201c;worker data&#x201d; that is locally-calculated, stored, and managed in individual worker computers. When a local computing process is completed, the &#x201c;worker data&#x201d; can be transferred to other worker computers so as to complete a whole computing process. The system enhances an overall computing efficiency and enables collaborative filtering across a large data set.

Collaborative filtering in directed graph

In one embodiment, a method includes receiving a number of notifications of activity relevant to a user. Each notification has an associated type. The method also includes calculating an estimated click-through rate (CTR) for each notification based at least in part on the type associated with each notification; determining a push threshold value for each notification based at least in part on the estimated CTR for each notification; and sending one or more of the notifications to the user. Each of the sent notifications has a push threshold value higher than a pre-determined push threshold value.

Notification click-through rate (CTR) thresholds

The disclosure is directed to evaluating feature vectors using decision trees. Typically, the number of feature vectors and the number of decision trees are very high, which prevents loading them into a processor cache. The feature vectors are evaluated by processing the feature vectors across a disjoint subset of trees repeatedly. After loading the feature vectors into the cache, they are evaluated across a first subset of trees, then across a second subset of trees and so on. If the values based on the first and second subsets satisfy a specified criterion, further evaluation of the feature vectors across the remaining of the decision trees is terminated, thereby minimizing the number of trees evaluated and therefore, consumption of computing resources.

Evaluating feature vectors across disjoint subsets of decision trees

In one embodiment, a method includes accessing for a user one or more electronic notifications that include information about a social network. The social network includes a plurality of nodes connected by a plurality of edges, with at least one node corresponding to the user. The method further includes determining, for each of the electronic notifications, a score that estimates whether a user interaction with the social network will result if the electronic notification is provided to the user. The method further includes determining, based at least in part on the determined scores, an action to take with respect to the electronic notifications.

Electronic notifications

Online systems generate predictors for predicting actions of users of the online system. The online system receives requests to generate predictor models for predicting whether a user is likely to take an action of a particular action type. The request specifies the type of action and criteria for identifying a successful instance of the action type and a failure instance of the action type. The online system collects data including successful and failure instances of the action type. The online system generates one or more predictors of different types using the generated data. The online system evaluates and compares the performance of the different predictors generated and selects a predictor based on the performance. The online system returns a handle to access the generated predictor to the requester of the predictor.

Growing community of inventors

Los Altos, CA, United States of America

Aleksandar Ilic