Information processing apparatus, information processing method, and program

A method for generating a similarity determination algorithm determining a similarity between a pair of data, the method comprising:

• generating a feature-quantity-extraction expression list including a plurality of feature-quantity-extraction expressions, by modifying the feature-quantity-extraction expression list of a preceding generation on the basis of an evaluation value of each of the feature-quantity-extraction expressions of a preceding generation, each of the feature-quantity-extraction expressions including an operator;
• inputting first teacher data and second teacher data into the feature-quantity-extraction expressions included in the feature-quantity-extraction expression list to calculate, for each of the feature-quantity-extraction expressions, a feature quantity corresponding to the first teacher data and a feature quantity corresponding to the second teacher data;
• calculating the evaluation value of each of the feature-quantity-extraction expressions included in the feature-quantity-extraction expression list using the calculated feature quantities and a known similarity between the first teacher data and the second teacher data;
• and
• estimating a similarity calculation expression using:
• a distance between feature quantities calculated by inputting the first teacher data and the second teacher data into one of the feature-quantity-extraction expressions, and
• the known similarity between the first teacher data and the second teacher data.

A non-transitory computer-readable medium storing instructions for generating a similarity determination algorithm determining a similarity between a pair of data, the instructions, when executed, causing a computer to:

• generate a feature-quantity-extraction expression list including a plurality of feature-quantity-extraction expressions, by modifying the feature-quantity-extraction expression list of a preceding generation on the basis of an evaluation value of each of the feature-quantity-extraction expressions of a preceding generation, each of the feature-quantity-extraction expressions including an operator;
• input first teacher data and second teacher data into the feature-quantity-extraction expressions included in the feature-quantity-extraction expression list to calculate, for each of the feature-quantity-extraction expressions, a feature quantity corresponding to the first teacher data and a feature quantity corresponding to the second teacher data;
• calculate the evaluation value of each of the feature-quantity-extraction expressions included in the feature-quantity-extraction expression list using the calculated feature quantities and a known similarity between the first teacher data and the second teacher data;
• and
• estimate a similarity calculation expression using:
• a distance between feature quantities calculated by inputting the first teacher data and the second teacher data into one of the feature-quantity-extraction expressions, and
• the known similarity between the first teacher data and the second teacher data.