Systems and methods for parameter optimization
Methods and systems that provide one or more recommended configurations to planners using large data sets in an efficient manner. These methods and systems provide optimization of objectives using a genetic algorithm that can provide parameter recommendations that optimize one or more objectives in an efficient and timely manner. The methods and systems disclosed herein are flexible enough to satisfy diverse use cases.
1. A computer-implemented method for parameter optimization, the method comprising the steps of:
defining, by a parameter optimization module, one or more objectives and one or more parameters;
generating, by the parameter optimization module, an initial set of values of the one or more parameters;
evaluating, by the parameter optimization module, a fitness function of each objective based on the set of values of the one or more parameters;
obtaining, by the parameter optimization module, a Pareto Front comprising Pareto objective points, each objective point associated with the fitness function;
applying, by the parameter optimization module, a genetic algorithm to the Pareto objective points to generate a new set of objective points;
evaluating, by the parameter optimization module, a new Pareto Front based on the new set of objective points;
comparing, by the parameter optimization module, a distance between successive Pareto Fronts; and
iterating, by the parameter optimization module, through successive Pareto Fronts until the distance is less than a threshold.
2. The method of claim 1 , wherein when applying the genetic algorithm, the method further comprises:
calculating, by the parameter optimization module, a plurality of crowding distances of the Pareto objective points; and
using, by the parameter optimization module, the plurality of crowding distances to determining pairings of the Pareto objective points.
3. The method of claim 1 , further comprising:
grouping, by the parameter optimization module, the one or more parameters as a tree structure, the tree structure comprising one or more node levels; and
applying, by the parameter optimization module, the genetic algorithm recursively to each node level of the tree structure.
4. A system for parameter optimization, the system comprising:
a processor; and
a memory storing instructions that, when executed by the processor, configure the system to:
define, by a parameter optimization module, one or more objectives and one or more parameters;
generate, by the parameter optimization module, an initial set of values of the one or more parameters;
evaluate, by the parameter optimization module, a fitness function of each objective based on the set of values of the one or more parameters;
obtain, by the parameter optimization module, a Pareto Front comprising Pareto objective points, each objective point associated with the fitness function;
apply, by the parameter optimization module, a genetic algorithm to the Pareto objective points to generate a new set of objective points;
evaluate, by the parameter optimization module, a new Pareto Front based on the new set of objective points;
compare, by the parameter optimization module, a distance between successive Pareto Fronts; and
iterate, by the parameter optimization module, through successive Pareto Fronts until the distance is less than a threshold.
5. The system of claim 4 , wherein during application of the genetic algorithm, the system is further configured to:
calculate, by the parameter optimization module, a plurality of crowding distances of the Pareto objective points;
use, by the parameter optimization module, the plurality of crowding distances to determine pairings of the Pareto objective points.
6. The system of claim 4 , wherein the system is further configured to:
group, by the parameter optimization module, the one or more parameters as a tree structure, the tree structure comprising one or more node levels; and
apply, by the parameter optimization module, the genetic algorithm recursively to each node level of the tree structure.
7. A non-transitory computer-readable storage medium, the computer-readable storage medium including instructions that when executed by a computer, cause the computer to:
define, by a parameter optimization module, one or more objectives and one or more parameters;
generate, by the parameter optimization module, an initial set of values of the one or more parameters;
evaluate, by the parameter optimization module, a fitness function of each objective based on the set of values of the one or more parameters;
obtain, by the parameter optimization module, a Pareto Front comprising Pareto objective points, each objective point associated with the fitness function;
apply, by the parameter optimization module, a genetic algorithm to the Pareto objective points to generate a new set of objective points;
evaluate, by the parameter optimization module, a new Pareto Front based on the new set of objective points;
compare, by the parameter optimization module, a distance between successive Pareto Fronts; and
iterate, by the parameter optimization module, through successive Pareto Fronts until the distance is less than a threshold.
8. The computer-readable storage medium of claim 7 , wherein when applying the genetic algorithm, the computer is further configured to:
calculate, by the parameter optimization module, a plurality of crowding distances of the Pareto objective points; and
use, by the parameter optimization module, the plurality of crowding distances to determining pairings of the Pareto objective points.
9. The computer-readable storage medium of claim 7 , wherein the computer is further configured to:
group, by the parameter optimization module, the one or more parameters as a tree structure, the tree structure comprising one or more node levels; and
apply, by the parameter optimization module, the genetic algorithm recursively to each node level of the tree structure.
10. A computer-implemented method for parameter optimization, the method comprising the steps of:
defining, by a parameter optimization module, one or more objectives and one or more parameters;
generating, by the parameter optimization module, a tree structure of the one or more parameters, the tree structure comprising a plurality of leaf nodes and one or more node levels;
generating, by the parameter optimization module, an initial population of trees;
evaluating, by the parameter optimization module, a fitness function of each objective at each leaf node of the plurality of leaf nodes;
obtaining, by the parameter optimization module, an initial Pareto Front comprising Pareto objective points, each objective point associated with the fitness function;
maintaining, by the parameter optimization module, a second plurality of leaf nodes associated with the initial Pareto Front;
applying recursively, by the parameter optimization module, a genetic algorithm to each node level in the tree structure of the leaf nodes that form the initial Pareto Front, until the initial Pareto Front converges to a final Pareto Front.
11. The method of claim 10 , wherein when applying the genetic algorithm recursively, the method further comprises:
calculating, by the parameter optimization module, a plurality of crowding distances of the leaf nodes and a distance of each parent node; and
using, by the parameter optimization module, the plurality of crowding distances and the distances of each parent node to determining pairings at each node level.
12. The method of claim 10 , wherein the genetic algorithm is a Non-dominated Sorting Genetic Algorithm (NSGA-II).
13. A system comprising:
a processor; and
a memory storing instructions that, when executed by the processor, configure the system to:
define, by a parameter optimization module, one or more objectives and one or more parameters;
generate, by the parameter optimization module, a tree structure of the one or more parameters, the tree structure comprising a plurality of leaf nodes and one or more node levels;
generate, by the parameter optimization module, an initial population of trees;
evaluate, by the parameter optimization module, a fitness function of each objective at each leaf node of the plurality of leaf nodes;
obtain, by the parameter optimization module, an initial Pareto Front comprising Pareto objective points, each objective point associated with the fitness function;
maintain, by the parameter optimization module, a second plurality of leaf nodes associated with the initial Pareto Front;
apply recursively, by the parameter optimization module, a genetic algorithm to each node level in the tree structure of the leaf nodes that form the initial Pareto Front, until the initial Pareto Front converges to a final Pareto Front.
14. The system of claim 13 , wherein when the genetic algorithm is applied recursively, the system is further configured to:
calculate, by the parameter optimization module, a plurality of crowding distances of the leaf nodes and a distance of each parent node; and
use, by the parameter optimization module, the plurality of crowding distances and the distances of each parent node to determining pairings at each node level.
15. The system of claim 13 , wherein the genetic algorithm is a Non-dominated Sorting Genetic Algorithm (NSGA-II).
16. A non-transitory computer-readable storage medium, the computer-readable storage medium including instructions that when executed by a computer, cause the computer to:
define, by a parameter optimization module, one or more objectives and one or more parameters;
generate, by the parameter optimization module, a tree structure of the one or more parameters, the tree structure comprising a plurality of leaf nodes and one or more node level;
generate, by the parameter optimization module, an initial population of trees;
evaluate, by the parameter optimization module, a fitness function of each objective at each leaf node of the plurality of leaf nodes;
obtain, by the parameter optimization module, an initial Pareto Front comprising Pareto objective points, each objective point associated with the fitness function;
maintain, by the parameter optimization module, a second plurality of leaf nodes associated with the initial Pareto Front;
apply recursively, by the parameter optimization module, a genetic algorithm to each node level in the tree structure of the leaf nodes that form the initial Pareto Front, until the initial Pareto Front converges to a final Pareto Front.
17. The computer-readable storage medium of claim 16 , wherein when applying the genetic algorithm recursively, the computer is further configured to:
calculate, by the parameter optimization module, a plurality of crowding distances of the leaf nodes and a distance of each parent node; and
use, by the parameter optimization module, the plurality of crowding distances and the distances of each parent node to determining pairings at each node level.
18. The computer-readable storage medium of claim 16 , wherein the genetic algorithm is a Non-dominated Sorting Genetic Algorithm (NSGA-II).