Facebook, Inc.

Justin Mitchell

David Harry Garcia

Samuel Odio

In one embodiment, a social networking system maintains a moving average of the number of connection problems, including socket timeouts and failed uploads, per client in a geographic area to determine whether the wireless data network serving the geographic area is overloaded. In response to detecting a network overload, the social networking system may transmit an instruction to the clients in the particular geographic area to enter one of a plurality of traffic throttling modes. In particular embodiments, the social networking system maintains a historical log of network overload conditions, and uses the historical log to generate an estimate of the wireless network capacity serving a geographic area. Thus, the social networking system may preemptively transmit instructions to clients to enter a bandwidth-conservation mode when the estimated traffic demand exceeds the estimated capacity for a particular geographic region.

Server-initiated bandwidth conservation policies

Methods, apparatuses and systems directed to detecting objects in user-uploaded multimedia such as photos and videos, determining the location at which the media was captured, inferring a set of users of a social network who were physically present at the time and place of capture, and pushing remarketing content to the set of inferred users for the detected objects, or alternatively, the competitors of the detected concepts.

Computer-vision content detection for connecting objects in media to users

In one embodiment, a system may receive sponsor specifications from a sponsor designating one or more concepts, monitor a plurality of activity streams for upload by users of matching multimedia objects, determine a geographic location associated with each matching multimedia object, receive a request from the sponsor associated with a specific geographic area, and send to a client device associated with the sponsor in response to the request information identifying the geographic locations associated with the respective matching multimedia objects that are within the specific geographic area.

Computer-vision content detection for sponsored stories

In particular embodiments, the newsfeed is rendered to a particular user using the locally stored copy of the resource, instead of pulling it back from the server-side. Thus, a user may see his or her content even before it is completely uploaded to the social networking system. In particular embodiments, the user may interact with the content, for example, by adding comments, captions, 'likes', etc. to the content even before it has been completely uploaded to the social networking system. In this manner, the user is provided an increased feeling of responsiveness that is independent of the current wireless link condition.

Generation of activity stream from server-side and client-side data

Methods, apparatuses and systems directed to sponsored story generation from an photo upload in an organic activity stream in a social networking site. A social networking system may apply computer image algorithms to detect image objects in user-uploaded images and videos, and promote them as sponsored stories.

Methods, apparatuses and systems directed to calculating a probability that a user or set of users will engage with a multimedia object for customizing content in a social networking system. In one embodiment, a generative model representing all users is utilized to calculate the probability that all users will interact with a given multimedia object. In another embodiment, the same generative model is utilized to calculate a user-specific engagement probability for a given multimedia object. In particular embodiments, the generative model uses Bayesian probability. In particular embodiments, one or more policies are applied to the photos to customize the browsing experience for users.

Growing community of inventors

Palo Alto, CA, United States of America

Justin Mitchell