Program generation apparatus and program generation method

A program generation method comprising:

• generating, by a computer, an image processing program by determining, based on genetic programming, a combination of partial programs to be included in the image processing program from a plurality of partial programs, the generating including evolution processing for evolving an individual formed by a combination of partial programs selected from the plurality of partial programs;
• recording, by the computer, information indicating a type of partial program that has been changed in number between pre- and post-evolution individuals and indicating difference between fitness levels calculated based on the pre- and post-evolution individuals, respectively, in a memory, each time the evolution processing is performed;
• and
• determining, by the computer, for each of the plurality of partial programs, a contribution level indicating how much an individual partial program contributes to increasing a fitness level, based on information accumulated in the memory within a predetermined time period and setting a selection probability for each of the plurality of partial programs in such a manner that a first partial program having a higher contribution level than a second partial program is given a higher setting value as a selection probability of the first partial program than the second partial program,
• wherein the evolution processing performed after the setting of the selection probability for each of the plurality of partial programs includes selecting a partial program to be included in a post-evolution individual in mutation processing from the plurality of partial programs according to selection probabilities set for respective ones of the plurality of partial programs.

A non-transitory computer-readable storage medium storing a computer program that causes a computer to perform a procedure comprising:

• generating an image processing program by determining, based on genetic programming, a combination of partial programs to be included in the image processing program from a plurality of partial programs, the generating including evolution processing for evolving an individual formed by a combination of partial programs selected from the plurality of partial programs;
• recording information indicating a type of partial program that has been changed in number between pre- and post-evolution individuals and indicating difference between fitness levels calculated based on the pre- and post-evolution individuals, respectively, in a memory, each time the evolution processing is performed;
• and
• determining, for each of the plurality of partial programs, a contribution level indicating how much an individual partial program contributes to increasing a fitness level, based on information accumulated in the memory within a predetermined time period and setting a selection probability for each of the plurality of partial programs in such a manner that a first partial program having a higher contribution level than a second partial program is given a higher setting value as a selection probability of the first partial program than the second partial program,
• wherein the evolution processing performed after the setting of the selection probability for each of the plurality of partial programs includes selecting a partial program to be included in a post-evolution individual in mutation processing from the plurality of partial programs according to selection probabilities set for respective ones of the plurality of partial programs.