Technologies for platform-targeted machine learning

Abstract

Technologies for platform-targeted machine learning include a computing device to generate a machine learning algorithm model indicative of a plurality of classes between which a user input is to be classified and translate the machine learning algorithm model into hardware code for execution on the target platform. Example instructions cause a processor to obtain dataset features indicative of a plurality of characteristics of an input dataset, rank, using multiple ranking algorithms, the dataset features, identify feature subsets for respective ones of the ranked dataset features, predict performance metrics based on the feature subsets, and select a final subset based on the predicted performance metrics.

1 . One or more machine-readable storage media comprising instructions that, when executed, cause at least one processor to at least:
 obtain dataset features indicative of a plurality of characteristics of an input dataset;   rank, using multiple ranking algorithms, the dataset features;   identify feature subsets for respective ones of the ranked dataset features;   predict performance metrics based on the feature subsets; and   select a final subset based on the predicted performance metrics.   
     
     
         2 . The one or more machine-readable storage media of  claim 1 , wherein the instructions, when executed, cause the at least one processor to determine whether a budget for feature extraction is reached. 
     
     
         3 . The one or more machine-readable storage media of  claim 2 , wherein the budget is a time budget. 
     
     
         4 . The one or more machine-readable storage media of  claim 2 , wherein the instructions, when executed, cause the at least one processor to generate, in response to a determination that the budget for feature extraction is not reached, a second performance metric based on the one or more performance metrics. 
     
     
         5 . The one or more machine-readable storage media of  claim 1 , wherein the instructions, when executed, cause the at least one processor to generate the performance metrics based on a cost metric indicative of an implementation cost of a corresponding feature. 
     
     
         6 . The one or more machine-readable storage media of  claim 1 , wherein the instructions, when executed, cause the at least one processor to evaluate the feature subsets using a proxy model to predict the performance metrics. 
     
     
         7 . The one or more machine-readable storage media of  claim 1 , wherein the instructions, when executed, cause the at least one processor to identify a feature subset for respective ones of the ranked dataset features including at least a threshold amount of ranked dataset features. 
     
     
         8 . The one or more machine-readable storage media of  claim 1 , wherein the performance metrics are predicted using a plurality of proxy models. 
     
     
         9 . A method for platform-targeted machine learning, the method comprising:
 obtaining dataset features indicative of a plurality of characteristics of an input dataset;   ranking, by executing an instruction with at least one processor, using multiple ranking algorithms, the dataset features;   identifying, by executing an instruction with the least one processor, a feature subset for respective ones of the ranked dataset features;   predicting, by executing an instruction with the least one processor, performance metrics based on the feature subsets; and   selecting, by executing an instruction with the least one processor, a final subset based on the predicted performance metric.   
     
     
         10 . The method of  claim 9 , further including determining whether a budget for feature extraction is reached. 
     
     
         11 . The method of  claim 10 , wherein the budget is a time budget. 
     
     
         12 . The method of  claim 10 , further including generating, in response to a determination that the budget for feature extraction is not reached, a second performance metric based on the one or more metrics. 
     
     
         13 . The method of  claim 9 , further including generating the performance metric based on a cost metric indicative of an implementation cost of a corresponding feature. 
     
     
         14 . The method of  claim 9 , further including evaluating the feature subsets using a proxy model to predict the performance metrics. 
     
     
         15 . The method of  claim 9 , further including identifying a feature subset for respective ones of the ranked dataset features containing at least a threshold amount of ranked dataset features. 
     
     
         16 . The method of  claim 9 , wherein the performance metrics are predicted using a plurality of proxy models. 
     
     
         17 . An apparatus comprising:
 at least one memory;   instructions in the apparatus; and   processor circuitry to execute the instructions to:
 obtain dataset features indicative of a plurality of characteristics of an input dataset; 
 rank, using multiple ranking algorithms, the dataset features; 
 identify feature subsets for respective ones of the ranked dataset features; 
 predict performance metrics based on the feature subsets; and 
 select a final subset based on the predicted performance metrics. 
   
     
     
         18 . The apparatus of  claim 17 , wherein the processor circuitry is to determine whether a budget for feature extraction is reached. 
     
     
         19 . The apparatus of  claim 18 , wherein the budget is a time budget. 
     
     
         20 . The apparatus of  claim 18 , wherein the processor circuitry is to generate a second performance metric based on the one or more metrics, in response to a determination that the budget for feature extraction is not reached. 
     
     
         21 . The apparatus of  claim 17 , wherein the processor circuitry is to generate the performance metric based on a cost metric indicative of an implementation cost of a corresponding feature. 
     
     
         22 . The apparatus of  claim 17 , wherein the processor circuitry is to evaluate the feature subsets using a proxy model to predict the performance metrics. 
     
     
         23 . The apparatus of  claim 17 , wherein the processor circuitry is to identify a feature subset for respective ones of the ranked dataset features containing at least a threshold amount of ranked dataset features. 
     
     
         24 . The apparatus of  claim 17 , wherein the processor circuitry is to predict the performance metrics using a plurality of proxy models.