Methods and apparatus to perform multi-level hierarchical demographic classification

Abstract

Methods and apparatus to perform multi-level hierarchical demographic classification are disclosed. An example apparatus includes a neural network structured to process inputs at an input layer to form first outputs at a first output layer representing first possible classifications of an individual according to a demographic classification system at a first hierarchical level, and to process the first outputs to form second outputs at a second output layer representing possible combined classifications of the individual corresponding to combinations of the first possible classifications and second possible classifications of the individual according to the classification system at a second different hierarchical level; and a selector to select one of the second outputs, and associate with the individual a respective one of the first possible classifications and a respective one of the second possible classifications corresponding to a respective one of the possible combined classifications represented by the selected second output.

1 - 20 . (canceled) 
     
     
         21 . A hierarchical classification apparatus comprising:
 interface circuitry;   machine readable instructions; and   processor circuitry to execute the machine readable instructions to at least:
 implement a neural network to process demographic information to classify an individual, the neural network including an input layer, a first output layer, a sorting layer and a second output layer, the input layer to accept the demographic information, the first output layer to output a first set of output values, ones of the first set of output values to correspond respectively with ones of a first set of demographic classifications associated with a first level of a classification hierarchy, the sorting layer to convert the first set of output values into a second set of output values that corresponds respectively with the first set of output values, the sorting layer to also sort the first set of output values into groups of sorted output values, ones of the groups to correspond respectively with ones of a second set of demographic classifications associated with a second level of the classification hierarchy different from the first level, the second output layer to convert the groups of sorted output values into a third set of output values, ones of the third set of output values to correspond respectively to ones of the groups; and 
 output a demographic classification for the individual based on the second set of output values and the third set of output values. 
   
     
     
         22 . The apparatus of  claim 21 , wherein the sorting layer is to perform softmax operations on the first set of output values to convert the first set of output values into the second set of output values. 
     
     
         23 . The apparatus of  claim 21 , wherein the second output layer is to perform max pooling operations on the groups of sorted output values to convert the groups of sorted output values into the third set of output values. 
     
     
         24 . The apparatus of  claim 21 , wherein the second output layer is to perform softmax operations on ones of the sorted output values selected from the groups of sorted output values to convert the groups of sorted output values into the third set of output values. 
     
     
         25 . The apparatus of  claim 21 , wherein the second set of output values represents probabilities that the individual belongs to the ones of the first set of demographic classifications. 
     
     
         26 . The apparatus of  claim 21 , wherein the third set of output values represents probabilities that the individual belongs to the ones of the second set of demographic classifications. 
     
     
         27 . The apparatus of  claim 21 , wherein the processor circuitry is to update the neural network based on the second set of output values and the third set of output values. 
     
     
         28 . At least one non-transitory computer readable medium comprising computer readable instructions to cause processor circuitry to at least:
 execute a neural network to process demographic information to classify an individual, the neural network including an input layer, a first output layer, a sorting layer and a second output layer, the input layer to accept the demographic information, the first output layer to output a first set of output values, ones of the first set of output values to correspond respectively with ones of a first set of demographic classifications associated with a first level of a classification hierarchy, the sorting layer to convert the first set of output values into a second set of output values that corresponds respectively with the first set of output values, the sorting layer to also sort the first set of output values into groups of sorted output values, ones of the groups to correspond respectively with ones of a second set of demographic classifications associated with a second level of the classification hierarchy different from the first level, the second output layer to convert the groups of sorted output values into a third set of output values, ones of the third set of output values to correspond respectively to ones of the groups; and   output a demographic classification for the individual based on the second set of output values and the third set of output values.   
     
     
         29 . The at least one non-transitory computer readable medium of  claim 28 , wherein the sorting layer is to perform softmax operations on the first set of output values to convert the first set of output values into the second set of output values. 
     
     
         30 . The at least one non-transitory computer readable medium of  claim 28 , wherein the second output layer is to perform max pooling operations on the groups of sorted output values to convert the groups of sorted output values into the third set of output values. 
     
     
         31 . The at least one non-transitory computer readable medium of  claim 28 , wherein the second output layer is to perform softmax operations on ones of the sorted output values selected from the groups of sorted output values to convert the groups of sorted output values into the third set of output values. 
     
     
         32 . The at least one non-transitory computer readable medium of  claim 28 , wherein the second set of output values represents probabilities that the individual belongs to the ones of the first set of demographic classifications. 
     
     
         33 . The at least one non-transitory computer readable medium of  claim 28 , wherein the third set of output values represents probabilities that the individual belongs to the ones of the second set of demographic classifications. 
     
     
         34 . The at least one non-transitory computer readable medium of  claim 28 , wherein the instructions are to cause the processor circuitry to update the neural network based on the second set of output values and the third set of output values. 
     
     
         35 . A hierarchical classification method comprising:
 executing a neural network to process demographic information to classify an individual, the neural network including an input layer, a first output layer, a sorting layer and a second output layer, the executing of the neural network including:
 accepting the demographic information at the input layer; 
 outputting a first set of output values from the first output layer, ones of the first set of output values to correspond respectively with ones of a first set of demographic classifications associated with a first level of a classification hierarchy; 
 converting, at the sorting layer, the first set of output values into a second set of output values that corresponds respectively with the first set of output values; 
 sorting, at the sorting layer, the first set of output values into groups of sorted output values, ones of the groups to correspond respectively with ones of a second set of demographic classifications associated with a second level of the classification hierarchy different from the first level; and 
 converting the groups of sorted output values into a third set of output values at the second output layer, ones of the third set of output values to correspond respectively to ones of the groups; and 
   outputting a demographic classification for the individual based on the second set of output values and the third set of output values.   
     
     
         36 . The method of  claim 35 , wherein the sorting layer performs softmax operations on the first set of output values to convert the first set of output values into the second set of output values. 
     
     
         37 . The method of  claim 35 , wherein the second output layer performs max pooling operations on the groups of sorted output values to convert the groups of sorted output values into the third set of output values. 
     
     
         38 . The method of  claim 35 , wherein the second output layer performs softmax operations on ones of the sorted output values selected from the groups of sorted output values to convert the groups of sorted output values into the third set of output values. 
     
     
         39 . The method of  claim 35 , wherein the second set of output values represents probabilities that the individual belongs to the ones of the first set of demographic classifications, and the third set of output values represents probabilities that the individual belongs to the ones of the second set of demographic classifications. 
     
     
         40 . The method of  claim 35 , further including updating the neural network based on the second set of output values and the third set of output values.