PARALLEL HYPOTHESIS FORMATION FOR MULTIPLE HYPOTHESES TRACKING

摘要

In an embodiment, a method finds an optimal set within collected data objects, each having a score and a list of incompatible objects. The method can include initializing a plurality of sets. Each set has compatible data objects. The method further includes raising the score of the plurality of sets by repeatedly adding compatible subsets of data objects to the plurality of sets. The plurality of sets have a higher score after adding the compatible subsets. The compatible subsets are created by (a) at each of multiple processors, generating a list of subsets of the collected data objects by loading the list of sets from a memory and testing data objects of the subsets for compatibility for each set, and (b) calculating a score for each subset in the list, each processor calculating the score for a particular subset in the generated list in parallel. The method further includes selecting a group of subsets having a highest score.

主权项

1. A method of finding an optimal set within collected data objects, each particular data object having a score and having a list of objects that are incompatible with that particular data object, the method comprising:
by a hardware processor: initializing a plurality of sets in a memory, each set having one or more compatible data objects; raise the score of the plurality of sets by repeatedly adding compatible subsets of data objects to the plurality of sets, the compatible subsets created by a plurality of processors by:
(a) at each of the plurality of processors, generating a list of subsets of the collected data objects by loading a list of sets from a memory and testing data objects of the subsets for compatibility for each set, and(b) calculating a score for each subset in the generated list, each one of the plurality of processors calculating the score for a particular subset in the generated list in parallel; and selecting a group of subsets having a highest score.