Abstract: Systems and methods for using social network information to predict complex phenomena. According to one embodiment the system or method comprises a Support Vector Machine classifier utilized to infer a pre-determined state of an individual, location, or event based on information gathered from a social network dataset. A conditional random field model can then be used to predict an individual's propensity toward that pre-determined state using features derived from the social network dataset. Performance of the conditional random field model can be enhanced by including features that are not only based on the status of net work connections, but are also based on the estimated encounters with individuals having the pre-determined state, including individuals other than network connections.

Abstract: Methods of crowdsourcing medical expert information are described. One such method includes receiving, from a first user during a first session over a network, a first query that includes a request for information regarding a patient condition, a therapy, and/or a medical test, and transmitting the query to a plurality of medical personnel. The first query further includes first session parameters selected by the first user from categories displayed by a processor to the first user. The method also includes receiving responses to the query from the plurality of medical personnel, and changing, based on the responses, an assigned probability of presenting to an index user, in response to a second query and during another session, at least one of a suggested diagnosis, a suggested therapy, a suggested inquiry to aid in establishing a diagnosis, a suggested medical test, a recommended source for further information, or a recommended medical professional.

Abstract: A method of categorizing the relationship of individuals from a collection of photos taken in multiple sets of events includes: searching the photo collection to identify individuals in the photo collection, determining the gender and age range of the identified individuals; and using the identified individuals, their occurrences and co-occurrences, their gender, and their ages to infer social relationships between the individuals based on a set of rules.

Abstract: A method of categorizing the relationship of individuals from a collection of photos taken in multiple sets of events includes: searching the photo collection to identify individuals in the photo collection, determining the gender and age range of the identified individuals; and using the identified individuals, their occurrences and co-occurrences, their gender, and their ages to infer social relationships between the individuals based on a set of rules.

Abstract: A technique for finding values which satisfy a set of constraints. The technique is used with local search procedures for finding such values and overcomes the tendency of such local search procedures to "get stuck" at local minima. The technique dynamically adds weight to constraints in the set which are not satisfied by the current set of values and uses the weights of the constraints to determine the next set of values to be used in the local search. In the disclosed embodiment, the technique is used in the GSAT greedy local search procedure. Also disclosed is a system for controlling a robot arm which uses the technique.

Abstract: Techniques for reducing the amount of junk e-mail received by a user of an e-mail system. A recipient description containing non-address information is added to an e-mail message. The user has an e-mail filter which has access to information which provides a model of the user. The e-mail filter uses the non-address information and the model information to determine whether the e-mail message should be provided to the user. The e-mail filter further has access to information which provides models of the user's correspondents. If the filter does not provide the message to the user, it uses the non-address information and the model information of the user's correspondents to determine who the message might be forwarded to. A sender of e-mail can also use the model information of the sender's correspondents together with the non-address information to determine who the message should be sent to. The techniques are used in a system for locating expertise.