Systems and methods for providing data-driven evolution of arbitrary data structures
View Patent ↗A system and process for generalizing an evolutionary process applied to a particular domain involving different problems includes a researcher module for generating a configuration specification applicable to a particular problem. An evolution module parses the configuration specification into a representative tree structure, assembles policies for each node in the tree structure, and generates candidate genomes using the policies for each node in the tree structure. The policies may be applied to new data or data from prior runs to generate candidate genomes. The evolution module translates internal representations of the generated candidate genomes into known representations of the candidate genome for evaluation in accordance with the particular domain parameters by a candidate evaluation module.
1. A method for facilitating evolutionary optimization of candidate genomes in a desired domain, wherein evolution of the candidate genomes and evaluation of the candidate genomes with domain specific data are performed securely from each other, the method:
receiving by an evolution module located in a first secure processing environment from a researcher module a configuration specification, wherein the researcher module automatically builds the configuration specification using defined rules;
parsing the configuration specification by the evolution module into a tree structure of specific coded data structures;
generating a policy assembly using the parsed tree structure of specific coded data structures to produce a representation tree of reproduction policy structures, wherein each instance of the reproduction policy structures represents one of a single leaf node or branch node in the representation tree;
creating by the evolution module an internal representation of a candidate genome in accordance with the representation tree of reproduction policy structures, wherein the reproduction policy structures are data-type specific;
translating by the evolution module the internal representation of the candidate genome to a known representation of the candidate genome, wherein the known representation of the candidate genome is known to a candidate evaluation module; and
providing the known representation of the candidate genome to the candidate evaluation module located in a second secure processing environment for fitness evaluation in the desired domain.
2. The method of claim 1 , wherein the evolution module creates a population of multiple candidate genomes and assigns a unique identifier to each of the multiple candidate genomes.
3. The method of claim 1 , wherein the reproduction policy structures are selected from the group consisting of data-type-specific reproduction operations and interpretive operations of lists of random floating point and integer numbers.
4. The method of claim 1 , wherein the evolution module creates the internal representation of a candidate genome using data from a previously generated candidate genome.
5. An evolution module in a first secure processing environment for generating populations of candidate genomes for evaluation by an independent candidate evaluation module located in a second secure processing environment in a specific domain, comprising:
a parser module for
receiving a configuration specification from a research module,
parsing the configuration specification into a tree structure of specific coded data structures, and
generating a policy assembly using the parsed tree structure of specific coded data structures to produce a representation tree of reproduction policy structures, wherein each instance of the reproduction policy structures represents one of a single leaf node or branch node in the representation tree; and further wherein the configuration specification is indicative of a known candidate genome representation that is recognizable to the independent candidate evaluation module; and
a population initialization module for initiating the creation of an internal representation of a candidate genome in accordance with the representation tree of reproduction policy structures, wherein the reproduction policy structures are data-type specific;
wherein the evolution module translates the internal representation of the candidate genome to the known representation of the candidate genome and provides the known representation of the candidate genome to the candidate evaluation module for fitness evaluation in the specific domain; and
further wherein, evolution of the candidate genomes and evaluation of the candidate genomes in the specific domain are performed securely from each other.
6. The evolution module of claim 5 , wherein the population initialization module assigns a unique identifier to each candidate genome.
7. The evolution module of claim 5 , wherein the reproduction policy structures are selected from the group consisting of data-type-specific reproduction operations and interpretive operations of lists of random floating point and integer numbers.