System and method for enriching datasets while learning

Abstract

Systems and methods for enriching datasets while learning are provided. For example, intermediate results of training machine learning algorithms may be obtained. Additional training examples may be selected based on the intermediate results. In some cases, synthetic examples may be generated based on the intermediate results. The machine learning algorithms may be further trained using the selected additional training examples and/or the generated synthetic examples.

1 - 20 . (canceled) 
     
     
         21 . A system for enriching datasets, the system comprising:
 at least one storage device configured to store at least one dataset comprising a plurality of training examples; and   at least one processor configured to:   train at least one machine learning algorithm using the plurality of training examples to obtain one or more intermediate results;   based on the one or more intermediate results, obtain at least one additional training example; and   train the at least one machine learning algorithm using the at least one additional training example.   
     
     
         22 . The system of  claim 21 , wherein training the at least one machine learning algorithm using the plurality of training examples generates at least one intermediate inference model, and the one or more intermediate results comprises outputs of the at least one intermediate inference model for the plurality of training examples. 
     
     
         23 . The system of  claim 21 , wherein the at least one dataset further comprises a plurality of validation examples, training the at least one machine learning algorithm using the plurality of training examples generates at least one intermediate inference model, and the one or more intermediate results comprises outputs of the at least one intermediate inference model for the plurality of validation examples. 
     
     
         24 . The system of  claim 21 , wherein the at least one processor is further configured to generate at least one synthetic example based on the one or more intermediate results to obtain the at least one additional training example. 
     
     
         25 . The system of  claim 21 , wherein the at least one processor is further configured to:
 transmit the one or more intermediate results to an external device using a communication device; and   in response to the transmission, receive the at least one additional training example.   
     
     
         26 . A method for enriching datasets, the method comprising:
 accessing at least one dataset, the at least one dataset comprises a plurality of training examples;   training at least one machine learning algorithm using the plurality of training examples to obtain one or more intermediate results;   based on the one or more intermediate results, obtaining at least one additional training example; and   training the at least one machine learning algorithm using the at least one additional training example.   
     
     
         27 . The method of  claim 26 , wherein obtaining at least one additional training example is further based on a quota requirement associated with the at least one additional training example. 
     
     
         28 . The method of  claim 26 , further comprising:
 determining that a quality of the at least one additional training example is below a selected threshold;   based on said determination, providing a request for a user to provide at least one new training example;   in response to the request, receiving from the user the at least one new training example; and   training the at least one machine learning algorithm using the at least one new training example.   
     
     
         29 . The method of  claim 26 , further comprising:
 based on the one or more intermediate results, determining that user intervention is required;   based on said determination, providing a request for the user to provide at least one new training example;   in response to the request, receiving from the user the at least one new training example; and   training the at least one machine learning algorithm using the at least one new training example.   
     
     
         30 . The method of  claim 26 , further comprising:
 based on the one or more intermediate results, selecting a user of a plurality of alternative users;   providing a request to the selected user to provide at least one new training example;   in response to the request, receiving from the user the at least one new training example; and   training the at least one machine learning algorithm using the at least one new training example.   
     
     
         31 . The method of  claim 26 , wherein training the at least one machine learning algorithm using the at least one additional training example comprises training the at least one machine learning algorithm using the plurality of training examples and the at least one additional training example. 
     
     
         32 . The method of  claim 26 , wherein training the at least one machine learning algorithm using the plurality of training examples generates at least one intermediate inference model, and training the at least one machine learning algorithm using the at least one additional training example comprises updating the intermediate inference model based on the at least one additional training example. 
     
     
         33 . The method of  claim 26 , wherein training the at least one machine learning algorithm using the plurality of training examples generates at least one intermediate inference model, and the one or more intermediate results comprises outputs of the at least one intermediate inference model for the plurality of training examples. 
     
     
         34 . The method of  claim 26 , wherein the at least one dataset further comprises a plurality of validation examples, training the at least one machine learning algorithm using the plurality of training examples generates at least one intermediate inference model, and the one or more intermediate results comprises outputs of the at least one intermediate inference model for the plurality of validation examples. 
     
     
         35 . The method of  claim 26 , wherein training the at least one machine learning algorithm using the plurality of training examples comprises at least one of minimizing a first objective function and maximizing a second objective function, and wherein the one or more intermediate results comprises at least one of a value of the first objective function and a value of the second objective function. 
     
     
         36 . The method of  claim 26 , further comprising selecting the at least one additional training example of a plurality of alternative training examples based on the one or more intermediate results. 
     
     
         37 . The method of  claim 26 , further comprising selecting at least one additional dataset of a plurality of alternative datasets based on the one or more intermediate results, the at least one additional dataset comprises the at least one additional training example. 
     
     
         38 . The method of  claim 26 , further comprising generating at least one synthetic example based on the one or more intermediate results to obtain the at least one additional training example. 
     
     
         39 . The method of  claim 26 , further comprising:
 transmitting the one or more intermediate results to an external device using a communication device; and   in response to the transmission, receiving the at least one additional training example.   
     
     
         40 . A non-transitory computer readable medium storing data and computer implementable instructions for carrying out a method for enriching datasets, the method comprising:
 accessing at least one dataset, the at least one dataset comprises a plurality of training examples;   training at least one machine learning algorithm using the plurality of training examples to obtain one or more intermediate results;   based on the one or more intermediate results, obtaining at least one additional training example; and   training the at least one machine learning algorithm using the at least one additional training example.