US 2017/0323219 A1Published application

Data mining technique with distributed novelty search

AssigneeSentient Technologies (Barbados) Limited

InventorsHormoz Shahrzad, Babak Hodjat, Risto Miikkulainen

Life of this patent

Estimating the term…

Loading patent…

Abstract

Roughly described, an evolutionary data mining system includes at least two processing units, each having a pool of candidate individuals in which each candidate individual has a fitness estimate and experience level. A first processing unit tests candidate individuals against training data, updates an individual's experience level, and assigns each candidate to one of multiple layers of the candidate pool based on the individual's experience level. Individuals within the same layer of the same pool compete with each other to remain candidates. The first processing unit selects a set of candidates to retain based on the relative novelty of their responses to the training data. The first processing unit reports successful individuals to the second processing unit, and receives individuals for further testing from the second processing unit. The second processing unit selects individuals to retain based on their fitness estimate.

Claims17

Dependent claims

An arrangement of processing units, for use with a data mining training database containing training data, comprising:

a first processing unit having a first processor and a first memory, the first memory storing a first local pool of candidate individuals, each of the candidate individuals in the first local pool identifying at least one classification rule and further identifying a respective fitness estimate centric to the first processing unit;
and
a second processing unit disposed down-chain of the first processing unit, the second processing unit having a second processor and a second memory, the second memory storing a second local pool of candidate individuals, each of the candidate individuals in the second local pool identifying at least one classification rule and further identifying a respective fitness estimate centric to the second processing unit,
wherein the first processor is configured to:
store in the first memory individuals for evaluation against portions of the training data,
delegate, to the second processing unit, evaluation of individuals from the first local pool of candidate individuals,
update the fitness estimates, centric to the first processing unit, for selected ones of individuals received back from the second processing unit after testing, in dependence upon results of such testing, and
select individuals for discarding from the first local pool, in dependence upon their fitness estimates relative to the fitness estimates of the other individuals within the first local pool,
and wherein the second processor is configured to:
store in the second local pool of candidate individuals, individuals received by the second processing unit from the first processing unit for evaluation,
test individuals from the second local pool of candidate individuals against a portion of the training data,
update the fitness estimates, centric to the second processing unit, for selected ones of individuals tested, in dependence upon the results of such testing,
select individuals for discarding from the second local pool, in dependence upon their novelty relative to other individuals in the second local pool, and
report to the first processing unit, individuals from the second local pool and not selected for discarding, in conjunction with the individuals' fitness estimates, centric to the second processing unit.

11 more elements

A computer system which transmits toward a second processing unit, software which, when executed by a second processor performs:

storing in a second local pool of candidate individuals, individuals received by the second processing unit from a first processing unit for evaluation,
testing individuals from the second local pool of candidate individuals against a portion of training data,
updating fitness estimates, centric to the second processing unit, for selected ones of individuals tested, in dependence upon results of such testing,
selecting individuals for discarding from the second local pool, in dependence upon their novelty relative to other individuals in the second local pool, and
reporting toward the first processing unit, individuals from the second local pool and not selected for discarding, in conjunction with the individuals' fitness estimates, centric to the second processing unit.

2 more elements

An evolutionary engine for use with a data mining training database containing training data, and for use further with an evolutionary coordinator having a pool of candidate individuals stored accessibly thereto, the evolutionary engine comprising:

a processor and a memory, the memory storing accessibly to the processor a local pool of candidate individuals, each of the candidate individuals in the local pool identifying at least one classification rule and further identifying a respective fitness estimate centric to the evolutionary engine,
wherein the processor is configured to:
store in the local pool, individuals received by the evolutionary engine from the evolutionary coordinator for evaluation,
test individuals from the local pool against a portion of the training data,
update the fitness estimates, centric to the evolutionary engine, for selected ones of individuals tested, in dependence upon the results of such testing,
select individuals for discarding from the local pool, in dependence upon their novelty relative to other individuals in the local pool, and
report toward the evolutionary coordinator individuals from the local pool and not selected for discarding, in conjunction with the individuals' fitness estimates centric to the evolutionary coordinator.

4 more elements

Dependencies are inferred from the claim text.