Search point determining method and search point determining apparatus

1. A search point determining method in an estimation process of a function, executed by a processor included in a search point determining apparatus, the method comprising:

calculating a search prediction time and a confidence interval upper limit obtained by using a Gaussian process for the function in each search candidate point from a first search result of the function;

generating an area in a parameter space for each search candidate point by using a position of a search point close to the relevant search candidate point in a first search result, a search prediction time corresponding to each search candidate point, and a confidence interval upper limit corresponding to each search candidate point;

determining a search point based on a value obtained by dividing a product of an improvement value from an evaluation value in the first search result at each search candidate point and a size of the area by the search prediction time in a plurality of parameter spaces;

when a remaining time until a time limit is longer than or equal to a certain threshold, outputting a second search result as a solution of the function by searching more widely than the first search result; and

when the remaining time is shorter than the certain threshold, outputting a second search result as the solution of the function by reusing the first search result.

2. The search point determining method according to claim 1 ,

wherein the area in the parameter space is an N-dimensional hyper-rectangular parallelepiped.

3. The search point determining method according to claim 1 ,

wherein the area in the parameter space is an N-dimensional hyperball.

4. The search point determining method according to claim 1 ,

wherein the generating corrects the size of the area in the parameter space according to the position of the search point in the first search result.

5. The search point determining method according to claim 1 ,

wherein the search candidate point is set within a range where an expected value of the function is higher than an evaluation value of a search point having the highest evaluation value in the first search result.

6. A search point determining apparatus that determines a search point in an estimation process of a function, the apparatus comprising:

a memory; and

a processor coupled to the memory and configured to:

calculate a search prediction time and a confidence interval upper limit obtained by using a Gaussian process for the function in each search candidate point from a first search result of the function,

generate an area in a parameter space for each search candidate point by using a position of a search point close to the relevant search candidate point in a first search result, a search prediction time corresponding to each search candidate point, and a confidence interval upper limit corresponding to each search candidate point,

determine a search point based on a value obtained by dividing a product of an improvement value from an evaluation value in the first search result at each search candidate point and a size of the area by the search prediction time in a plurality of parameter spaces,

when a remaining time until a time limit is longer than or equal to a certain threshold, output a second search result as a solution of the function by searching more widely than the first search result, and

when the remaining time is shorter than the certain threshold, output a second search result as the solution of the function by reusing the first search result.

7. The apparatus according to claim 6 ,

wherein the area in the parameter space is an N-dimensional hyper-rectangular parallelepiped.

8. The apparatus according to claim 6 ,

wherein the area in the parameter space is an N-dimensional hyperball.

9. The apparatus according to claim 6 ,

wherein the processor is configured to generate corrects the size of the area in the parameter space according to the position of the search point in the first search result.

10. The apparatus according to claim 6 ,

wherein the processor is configured to set the search candidate point within a range where an expected value of the function is higher than an evaluation value of a search point having the highest evaluation value in the first search result.

11. A non-transitory computer-readable recording medium storing a program that causes a processor included in a search point determining apparatus to execute a process, the process comprising:

calculating a search prediction time and a confidence interval upper limit obtained by using a Gaussian process for the function in each search candidate point from a first search result of the function;

generating an area in a parameter space for each search candidate point by using a position of a search point close to the relevant search candidate point in a first search result, a search prediction time corresponding to each search candidate point, and a confidence interval upper limit corresponding to each search candidate point;

determining a search point based on a value obtained by dividing a product of an improvement value from an evaluation value in the first search result at each search candidate point and a size of the area by the search prediction time in a plurality of parameter spaces;

when a remaining time until a time limit is longer than or equal to a certain threshold, outputting a second search result as a solution of the function by searching more widely than the first search result; and

when the remaining time is shorter than the certain threshold, outputting a second search result as the solution of the function by reusing the first search result.