Systems and methods for event prediction using schema networks

1. A system for event prediction using schema networks comprises:

a first antecedent entity state that represents a first entity at a first time; wherein the first antecedent entity state comprises a first assignment of values to an attribute set of the first entity;

a first consequent entity state that represents the first entity at a second time; wherein the first consequent entity state comprises a second assignment of values to the attribute set of the first entity;

a second antecedent entity state that represents a second entity at the first time; wherein the second antecedent entity state comprises a first assignment of values to an attribute set of the second entity; and

a first schema factor that couples the first and second antecedent entity states to the first consequent entity state; wherein the first schema factor is configured to predict the first consequent entity state from the first and second antecedent entity states; wherein the system is implemented as a network of electronically coupled functional node components by a computer processor;

wherein the first schema factor is additionally configured to explain a cause of a transition from the first antecedent entity state to the first consequent entity state;

wherein the first schema factor is configured to explain the cause of a transition from the first antecedent entity state to the first consequent entity state using backward message propagation from the first consequent entity state to the first antecedent entity state.

2. The system of claim 1 , wherein the first schema factor is an AND factor node that performs prediction using one of max-product and sum-product belief propagation.

3. A system for event prediction using schema networks comprises:

a first antecedent entity state that represents a first entity at a first time; wherein the first antecedent entity state comprises a first assignment of values to an attribute set of the first entity;

a first consequent entity state that represents the first entity at a second time; wherein the first consequent entity state comprises a second assignment of values to the attribute set of the first entity;

a second antecedent entity state that represents a second entity at the first time; wherein the second antecedent entity state comprises a first assignment of values to an attribute set of the second entity; and

a first schema factor that couples the first and second antecedent entity states to the first consequent entity state; wherein the first schema factor is configured to predict the first consequent entity state from the first and second antecedent entity states; wherein the system is implemented as a network of electronically coupled functional node components by a computer processor;

wherein the first schema factor is additionally configured to use regression planning to determine at least one value of the first and second antecedent entity states based on given values of the first consequent entity state.

4. The system of claim 3 , wherein the first schema factor is configured to use regression planning to determine at least one value of the first and second antecedent entity states based on given values of the first consequent entity state using backward message propagation from the first consequent entity state to the first antecedent entity state.

5. A system for event prediction using schema networks comprises:

a first antecedent entity state that represents a first entity at a first time; wherein the first antecedent entity state comprises a first assignment of values to an attribute set of the first entity;

a first consequent entity state that represents the first entity at a second time; wherein the first consequent entity state comprises a second assignment of values to the attribute set of the first entity;

a second antecedent entity state that represents a second entity at the first time; wherein the second antecedent entity state comprises a first assignment of values to an attribute set of the second entity; and

a first schema factor that couples the first and second antecedent entity states to the first consequent entity state; wherein the first schema factor is configured to predict the first consequent entity state from the first and second antecedent entity states; wherein the system is implemented as a network of electronically coupled functional node components by a computer processor;

wherein the attribute set of the first entity comprises a first observable variable and a first latent variable; wherein the first schema factor is configured to predict a value of the first observable variable in the first consequent entity state based on values of the first observable variable and the first latent variable in the first antecedent entity state, and at least one variable of the attribute set of the second entity in the second antecedent entity state.

6. The system of claim 5 , wherein the second time is identical to the first time.

7. The system of claim 5 , wherein the second time is simultaneous with the first time.

8. The system of claim 5 , wherein the first observable variable corresponds to an observed attribute of an object identified by a visual network; wherein the first latent variable corresponds to a non-directly-observed physical characteristic of the object.

9. The system of claim 8 , wherein the object is a rubber ball, the observed attribute is a position of the rubber ball, and the non-directly-observed physical characteristic of the object is a hardness of the rubber ball.

10. The system of claim 8 , wherein the visual network is a recursive cortical network and both forward and backward message passing is enabled between the visual network and the system.

11. The system of claim 5 , further comprising an entity dictionary that encodes relationships between attribute sets of entities of the system.

12. The system of claim 11 , wherein the entity dictionary encodes a hierarchy of entities.

13. The system of claim 12 , wherein the first observable variable corresponds to an observed attribute of an object identified by a visual network; wherein the hierarchy of entities is determined by a hierarchy of the visual network.

14. A system for event prediction using schema networks comprises:

a first antecedent entity state that represents a first entity at a first time; wherein the first antecedent entity state comprises a first assignment of values to an attribute set of the first entity;

a first consequent entity state that represents the first entity at a second time; wherein the first consequent entity state comprises a second assignment of values to the attribute set of the first entity;

a second antecedent entity state that represents a second entity at the first time; wherein the second antecedent entity state comprises a first assignment of values to an attribute set of the second entity;

a first schema factor that couples a first subset of variables of the first and second antecedent entity states to a first schema variable;

a second schema factor that couples a second subset of variables of the first and second antecedent entity states to a second schema variable; and

a schema OR factor that couples the first and second schema variables to a variable of the first consequent entity state;

wherein the schema OR factor is configured to predict the first consequent entity state from the first and second antecedent entity states using the first and second schema factors;

wherein the system is implemented as a network of electronically coupled functional node components by a computer processor.

15. The system of claim 14 , wherein the first and second schema factors are AND factor nodes that perform prediction using one of max-product and sum-product belief propagation; wherein the schema OR factor is an OR factor node.

16. The system of claim 15 , wherein the schema OR factor is configured to explain a cause of a transition from the first antecedent entity state to the first consequent entity state using backward message propagation; wherein the cause of the transition arises from either the first schema factor or the second schema factor.

17. The system of claim 14 , further comprising a self-transition factor that couples a third subset of variables of the first antecedent entity state to a self-transition variable; wherein the schema OR factor couples the first and second schema variables and the self-transition variable to the variable of the first consequent entity state; wherein the self-transition factor and variable model the probability of a state change occurring in absence of any state change predicted by a schema factor.