edu.rice.hj.experimental.api

Interface HjExtremaEureka<T>

All Superinterfaces:

HjEureka<T,T,T>

All Known Implementing Classes:

HjExtremaEurekaImpl

public interface HjExtremaEureka<T>
extends HjEureka<T,T,T>

This Eureka tries to maximize its current value.

Many problems from artificial intelligence can be defined as combinatorial optimization problems. Subproblems are derived from the originally given problem through the addition of new constraints. The structure of the algorithm requires the ability to terminate individual subtrees of the search tree. An example is where we are interested in finding the lowest index of the goal item if it exists in the array.

The GUB is available in the HjExtremaEureka instance, eu, that a speculative task is registered on and can be retrieved by a call to eu.get(). Calls to offer() and check() pass the current known bounds or solution, respectively, as the argument. If the argument in the offer() call is lower than the GUB, the GUB is updated in the HjExtremaEureka instance, otherwise the current task is terminated. Similarly, calls to check() terminate a task if the argument is larger than the currently known GUB in eu.

The first argument of used in the comparator is the existing value in the eureka, and the second argument is the new argument passed inside the call to offer and check.

Instances are created using a factory method in the HJlib runtime, e.g., ModuleZ.newExtremaEureka(Object, java.util.Comparator)

Author:

Shams Imam (shams@rice.edu)

Method Summary

Methods inherited from interface edu.rice.hj.experimental.api.HjEureka

check, check, get, offer, offer, onCleanup, onRegistration, resolved, shouldAbortOnOffer