Method for determining class of data to be determined using machine learning model, information processing device, and computer program

1 . A method for determining a class of data to be determined using a vector neural network type machine learning model having a plurality of vector neuron layers, the method comprising:

step (a), preparing, for each of a plurality of classes, a known feature spectrum group obtained from an output of a specific layer among the plurality of vector neuron layers when a plurality of pieces of training data are input to the machine learning model; and

step (b), executing, by using the machine learning model and the known feature spectrum group, a class determination processing on the data to be determined, wherein

the step (b) includes

step (b1), calculating a feature spectrum based on the output of the specific layer according to an input of the data to be determined to the machine learning model,

step (b2), calculating a class similarity between the feature spectrum and the known feature spectrum group related to each of the plurality of classes, and

step (b3), determining a class of the data to be determined according to the class similarity,

wherein:

the data to be determined is measurement data on spectral reflectance of a print medium, which is acquired by a spectrometer,

the specific layer includes two or more vector neuron layers, each of the vector neuron layers in the specific layer has a configuration in which vector neurons are arranged on a conceptual plane defined by two axes of a first axis and a second axis and are arranged as a plurality of channels along a third axis in a depth direction intersecting with the two axes,

in each vector neuron layer of the specific layer, when a region specified by a plane position defined by a position of the first axis and a position of the second axis and including the plurality of channels along the third axis is referred to as a partial region,

for each partial region of a plurality of partial regions included in each vector neuron layer of the specific layer, the feature spectrum is obtained as any one of

(i) a first type of feature spectrum obtained by arranging a plurality of element values of an output vector of each vector neuron included in the partial region over the plurality of channels along the third axis,

(ii) a second type of feature spectrum obtained by multiplying each element value of the first type of feature spectrum by a normalization coefficient corresponding to a vector length of the output vector, and

(iii) a third type of feature spectrum obtained by arranging the normalization coefficient over the plurality of channels along the third axis, and

the step (b2) includes

obtaining a plurality of local similarities indicating similarities for each class related to the plurality of partial regions in each vector neuron layer of the specific layer by obtaining a local similarity which is a similarity between the feature spectrum obtained from an output of a specific partial region in each vector neuron layer of the specific layer according to the data to be determined and all the known feature spectra associated with each vector neuron layer of the specific layer and each class, and

obtaining, for each class, the class similarity by taking a maximum value, an average value, or a minimum value of the plurality of local similarities related to the plurality of partial regions.

2 . The method according to claim 1 , wherein all the known feature spectra used in the step (b2) is associated with the specific partial region in each vector neuron layer of the specific layer.

3 . The method according to claim 1 , wherein

when there are a plurality of the specific layers,

the step (b3) includes

obtaining respective class determination results related to the plurality of specific layers by obtaining a class determination result based on the class similarity related to each vector neuron layer of each specific layer, and

determining, among classes indicated by the class determination results, a class that appears most frequently as a determined class of the data to be determined.

4 . The method according to claim 1 , wherein when there are a plurality of the specific layers,

the step (b3) includes

determining, for each partial region in each vector neuron layer of each specific layer, a class based on the local similarity related to the partial region, and assigning, to the partial region, a class parameter value indicating an order of the class among the plurality of classes,

calculating, for each vector neuron layer of each of the plurality of specific layers, a statistical variance related to a distribution of the class parameter values in the plurality of partial regions, and

determining a class of the data to be determined using the class similarity obtained for a vector neuron layer of a specific layer having the smallest variance among the plurality of specific layers.

5 . The method according to claim 1 , wherein

the determined class of the data indicates a type of the print medium, and

a printing condition of a printer is adjusted based on the determined type of the print medium.

6 . The method according to claim 1 , wherein the step (b2) includes obtaining, for each class, three types of the class similarity by taking, respectively, a maximum value, an average value, and a minimum value of the plurality of local similarities related to the plurality of partial regions.

7 . The method according to claim 1 , wherein an output of each of the vector neuron layers in the specific layer is a vector where spatial information of the feature spectrum including a direction and a scale thereof is preserved.

8 . An information processing device that executes a class determination processing of determining a class of data to be determined using a vector neural network type machine learning model having a plurality of vector neuron layers, the information processing device comprising:

a memory that stores the machine learning model; and

a processor that executes a calculation using the machine learning model, wherein

the processor is configured to execute

a processing (a) of, reading, from the memory for each of the plurality of classes, a known feature spectrum group obtained from an output of a specific layer among the plurality of vector neuron layers when a plurality of pieces of training data are input to the machine learning model, and

a processing (b) of executing, by using the machine learning model and the known feature spectrum group, a class determination processing on the data to be determined, and

the processing (b) includes

a processing (b1) of calculating a feature spectrum based on an output of the specific layer according to an input of the data to be determined to the machine learning model,

a processing (b2) of calculating a similarity between the feature spectrum and the known feature spectrum group related to each of the plurality of classes, and

a processing (b3) of determining a class of the data to be determined according to the similarity,

wherein:

the data to be determined is measurement data on spectral reflectance of a print medium, which is acquired by a spectrometer,

the specific layer includes two or more vector neuron lavers, each of the vector neuron lavers in the specific layer has a configuration in which vector neurons are arranged on a conceptual plane defined by two axes of a first axis and a second axis and are arranged as a plurality of channels along a third axis in a depth direction intersecting with the two axes,

in each vector neuron layer of the specific layer, when a region specified by a plane position defined by a position of the first axis and a position of the second axis and including the plurality of channels along the third axis is referred to as a partial region,

for each partial region of a plurality of partial regions included in each vector neuron layer of the specific layer, the feature spectrum is obtained as any one of

(i) a first type of feature spectrum obtained by arranging a plurality of element values of an output vector of each vector neuron included in the partial region over the plurality of channels along the third axis,

(ii) a second type of feature spectrum obtained by multiplying each element value of the first type of feature spectrum by a normalization coefficient corresponding to a vector length of the output vector, and

(iii) a third type of feature spectrum obtained by arranging the normalization coefficient over the plurality of channels along the third axis, and

the processing (b2) includes

obtaining a plurality of local similarities indicating similarities for each class related to the plurality of partial regions in each vector neuron layer of the specific layer by obtaining a local similarity which is a similarity between the feature spectrum obtained from an output of a specific partial region in each vector neuron layer of the specific layer according to the data to be determined and all the known feature spectra associated with each vector neuron layer of the specific layer and each class, and

obtaining, for each class, the class similarity by taking a maximum value, an average value, or a minimum value of the plurality of local similarities related to the plurality of partial regions.

9 . A non-transitory computer-readable storage medium storing a computer program for causing a processor to execute a class determination processing of determining a class of data to be determined using a vector neural network type machine learning model having a plurality of vector neuron layers, wherein

the computer program causes the processor to execute

a processing (a) of reading, from a memory for each of a plurality of classes, a known feature spectrum group obtained from an output of a specific layer among the plurality of vector neuron layers when a plurality of pieces of training data are input to the machine learning model, and

a processing (b) of executing, by using the machine learning model and the known feature spectrum group, a class determination processing on the data to be determined, and

the processing (b) includes

a processing (b1) of calculating a feature spectrum based on an output of the specific layer according to an input of the data to be determined to the machine learning model,

a processing (b2) of calculating a similarity between the feature spectrum and the known feature spectrum group related to each of the plurality of classes, and

a processing (b3) of determining a class of the data to be determined according to the similarity,

wherein:

the data to be determined is measurement data on spectral reflectance of a print medium, which is acquired by a spectrometer,

the specific layer includes two or more vector neuron layers, each of the vector neuron lavers in the specific layer has a configuration in which vector neurons are arranged on a conceptual plane defined by two axes of a first axis and a second axis and are arranged as a plurality of channels along a third axis in a depth direction intersecting with the two axes,

in each vector neuron layer of the specific layer, when a region specified by a plane position defined by a position of the first axis and a position of the second axis and including the plurality of channels along the third axis is referred to as a partial region,

for each partial region of a plurality of partial regions included in each vector neuron layer of the specific layer, the feature spectrum is obtained as any one of

(i) a first type of feature spectrum obtained by arranging a plurality of element values of an output vector of each vector neuron included in the partial region over the plurality of channels along the third axis,

(ii) a second type of feature spectrum obtained by multiplying each element value of the first type of feature spectrum by a normalization coefficient corresponding to a vector length of the output vector, and

(iii) a third type of feature spectrum obtained by arranging the normalization coefficient over the plurality of channels along the third axis, and

the processing (b2) includes

obtaining a plurality of local similarities indicating similarities for each class related to the plurality of partial regions in each vector neuron layer of the specific layer by obtaining a local similarity which is a similarity between the feature spectrum obtained from an output of a specific partial region in each vector neuron layer of the specific layer according to the data to be determined and all the known feature spectra associated with each vector neuron layer of the specific layer and each class, and

obtaining, for each class, the class similarity by taking a maximum value, an average value, or a minimum value of the plurality of local similarities related to the plurality of partial regions.