CPC G06F 11/3688 (2013.01) [G06F 11/3684 (2013.01); G06N 5/046 (2013.01)] | 14 Claims |
1. A software test method to be executed by an information processing device, the method comprising:
processing of storing, in a storage device, a prediction model generated by machine learning, the prediction model being a test target;
processing of accepting an input of a precondition for determining a range of a test input value;
processing of accepting an input of a constraint condition for an inference output value of a prediction model with respect to the test input value;
processing of accepting an input of an approximation threshold value;
processing of acquiring the prediction model from the storage device and converting the prediction model into a logical expression;
processing of analyzing an approximation range based on the approximation threshold value with respect to the logical expression to simplify the logical expression;
processing of generating an inspection expression by combining the simplified logical expression with the precondition and negation of the constraint condition;
processing of searching for, as a counterexample, a value satisfying the inspection expression;
processing of inputting the value to the prediction model to evaluate inspection accuracy when the counterexample exists; and
processing of outputting a result of the evaluation.
|