Learning device, estimating device, learning method, and computer program product

A learning method, comprising:

• calculating a likelihood of input data belonging to a plurality of classes of learning data, by using an estimation model to estimate which of the estimated classes the data belongs;
• calculating a weight of a loss function to be used in learning the estimation model, such that when a likelihood of a first class that is closer to the input data than other estimated classes among the estimated classes, and likelihoods of a second class and a third class that are adjacent to the first class are applied to a function comprising a particular shape function, a position that is a minimum value or maximum value of the particular shape function corresponds to the input data;
• and
• learning the estimation model by using the loss function.

A computer program product comprising a non-transitory computer readable medium comprising programmed instructions, wherein the instructions, when executed by a computer, cause the computer to perform:

• calculating a likelihood of input data belonging to a plurality of classes of learning data, by using an estimation model to estimate which of the estimated classes the input data belongs;
• calculating a weight of a loss function to be used in learning the estimation model such that, when a likelihood of a first class that is closer to the input data than other estimated classes among the estimated classes and likelihoods of a second class and a third class that are adjacent to the first class are applied to a function comprising a particular shape function, a position that is a minimum value or maximum value of the particular shape function corresponds to the input data;
• and
• learning the estimation model by using the loss function.