Affinity propagation in adaptive network-based systems
Adaptive applications of affinity propagation are described to facilitate effective and computationally efficient means of clustering computer-based objects such as items of content, and/or to determine exemplars associated with a set of objects. Affinity propagation is also applied by the present invention to define system user affinity groups and/or exemplar users. The present invention applies usage behaviors as a basis for influencing clustering through methods such as initializing exemplar attractor values based on usage behaviors and/or basing similarity values between pairs of objects or users on usage behaviors associated with system objects, or usage behaviors that are associated with, directly or indirectly, specific system users.
1. An apparatus, comprising:
a structural aspect comprising objects;
a usage aspect comprising usage behaviors that correspond to usage behavior categories; and
logic circuitry configured to:
identify affinities between the objects based on the usage behaviors;
derive an influence metric based on the affinities between the objects; and
identify a subset of the objects based on the influence metric;
wherein the influence metric comprises first degree and second degree influences.
2. The apparatus of claim 1 , wherein the affinities have different degrees of separation from the objects and the influence metric is derived based on the different degrees of separation.
3. The apparatus of claim 1 , wherein the logic circuitry is configured to initialize exemplar attractor values for the objects.
4. The apparatus of claim 3 , wherein the logic circuitry is configured to identify one of the objects as an exemplar.
5. The apparatus of claim 4 , wherein the logic circuitry is configured to provide the exemplar as a recommendation.
6. The apparatus of claim 3 , wherein the logic circuitry is configured to initialize similarity values between the objects based on the usage behaviors.
7. The apparatus of claim 3 , wherein the structural aspect comprises:
relationships between the objects; and
relationship indicators associated with the relationships.
8. The apparatus of claim 7 , wherein the logic circuitry is configured to initialize similarity values between the objects based on the relationship indicators.
9. The apparatus of claim 7 , wherein the relationship indicators are based on affinity propagation message values associated with the objects.
10. An apparatus, comprising:
a structural aspect comprising objects;
a usage aspect comprising a plurality of usage behaviors that correspond to a plurality of usage behavior categories; and
a computing device configured to:
identify affinities between the objects based on the plurality of usage behaviors;
derive an influence metric based on the affinities between the objects; and
identify a subset of the objects based on the influence metric;
wherein the influence metric comprises first degree and second degree influences.
11. The apparatus of claim 10 ,
wherein the plurality of usage behaviors are associated with users and the plurality of usage behavior categories identify different types of interactions of the users with the objects; and
wherein the computing device is configured to generate the influence metric based on different degrees of separation of the affinities from the users.
12. The apparatus of claim 11 , wherein the computing device is configured to identify the affinities between the users based on affinity propagation message values associated with the users.
13. The apparatus of claim 10 , wherein the computing device is configured to initialize exemplar attractor values based on the influence metric.
14. A method, comprising:
accessing, by a computing device, a structural aspect comprising objects;
accessing, by the computing device, a usage aspect comprising usage behaviors that correspond to a plurality of usage behavior categories; and
identifying, by the computing device, a subset of the structural aspect by initializing exemplar attractor values for the objects based on an influence metric derived from the plurality of usage behaviors;
wherein the influence metric comprises first degree and second degree influences.
15. The method of claim 14 , further comprising:
identifying, by the computing device, affinities between the objects based on the usage behaviors; and
deriving, by the computing device, the influence metric based on degrees of separation of the affinities from the objects.
16. The method of claim 14 , wherein said identifying a subset of the structural aspect comprises identifying one of the objects in the subset of the structural aspect as an exemplar.
17. The method of claim 14 , wherein the plurality of usage behavior categories identify different types of interactions of users with the objects.
18. The method of claim 14 , further comprising initializing, by the computing device, similarity values between the objects based on the usage behaviors.
19. The method of claim 14 , further comprising initializing, by the computing device, similarity values between the objects based on relationship indicators associated with the objects.
20. The method of claim 14 , further comprising determining, by the computing device, a relationship indicator for a pair of the objects based on an affinity propagation message value associated with the pair of the objects.
21. A method, comprising:
identifying, by a computing device, a structural aspect comprising objects;
identifying, by the computing device, a usage aspect comprising usage behaviors that correspond to usage behavior categories;
identifying, by the computing device, affinities between the objects based on the usage behaviors;
deriving, by the computing device, an influence metric based on the affinities between the objects; and
identifying, by the computing device, a subset of the objects based on the influence metric;
wherein the influence metric comprises first degree and second degree influences.
22. The method of claim 21 , further comprising identifying, by the computing device, a subset of the users based on the influence metric.
23. The method of claim 21 , wherein the influence metric is based on the usage behavior categories.
24. The method of claim 23 wherein the usage behavior categories are associated with different types of user accesses to the objects and different types of user interactions with the objects.
25. The method of claim 21 , further comprising initializing, by the computing device, exemplar attractor values for the objects.
26. The method of claim 25 , further comprising identifying, by the computing device, exemplars based on the exemplar attractor values.
27. The method of claim 26 , further comprising initializing, by the computing device, similarity values between the objects.
28. The method of claim 26 , further comprising using, by the computing device, the exemplars as user recommendations.
29. The method of claim 21 , wherein the structural aspect comprises:
relationships between the objects; and
relationship indicators associated with the relationships.
30. The method of claim 29 , wherein the relationship indicators are associated with affinity propagation message values exchanged between the objects.