Systems and methods to use neural networks for model transformations

Abstract

Systems and methods for transforming legacy models and transforming a model into a neural network model are disclosed. In an embodiment, a method may include receiving input data comprising an input model, an input dataset, and an input command. The method may include applying the input model to the input dataset to generate model output and storing model output and at least one of input model features or a map of the input model. The method may include generating a candidate neural network models with parameters. The method may include tuning the candidate neural network models to the input model. The method may include receiving model output from the candidate neural network models and selecting a neural network model from the candidate neural network models based on the candidate model output and the model selection criteria. In some aspects, the method may include returning the selected neural network model.

1 - 20 . (canceled) 
     
     
         21 . A system for transforming an input model into a neural network model, the system comprising:
 one or more memory units for storing instructions; and   one or more processors configured to execute the instructions to perform operations comprising:
 receiving input data comprising the input model; 
 obtaining dataset model output by applying the input model to an input dataset; 
 storing the generated dataset model output and at least one of: 
 input model features; or a map of the input model features; 
 generating a plurality of candidate neural network models based on the input model features; 
 tuning the plurality of candidate neural network models by adjusting at least one of a plurality of hidden layers, a plurality of inputs, or a type of layer during tuning such that at least one of the input model features is reproduced; 
 selecting a neural network model from the plurality of candidate neural network models based on one or more model selection criteria; and 
 returning the selected neural network model. 
   
     
     
         22 . The system of  claim 21 , wherein the input model comprises a regression model. 
     
     
         23 . The system of  claim 21 , wherein the input dataset comprises synthetic data being a representation of original data. 
     
     
         24 . The system of  claim 21 , wherein obtaining the dataset model output comprises:
 sending a command to a development instance to apply the input model to the input dataset; and   obtaining the dataset model output from the development instance.   
     
     
         25 . The system of  claim 24 , wherein the development instance comprises a virtual machine or an ephemeral container instance. 
     
     
         26 . The system of  claim 21 , wherein generating the plurality of candidate neural network models comprises sending a command to a development instance to generate the plurality of candidate neural network models based on the input model features or the map of the input model features. 
     
     
         27 . The system of  claim 26 , wherein the development instance comprises a virtual machine or an ephemeral container instance. 
     
     
         28 . The system of  claim 26 , wherein:
 the command specifies one or more model parameters; and   generating the plurality of neural network models comprises generating at least one of the plurality of neural network models based on the one or more model parameters.   
     
     
         29 . The system of  claim 26 , wherein:
 the command specifies a model type; and   generating the plurality of candidate neural network models comprises generating at least one candidate neural network model of the specified model type.   
     
     
         30 . The system of  claim 29 , wherein:
 the command further specifies a number of candidate neural network models of the specified model type; and   generating the plurality of candidate neural network models comprises generating the specified number of candidate neural network models of the specified type.   
     
     
         31 . The system of  claim 21 , wherein generating the plurality of candidate neural network models comprises retrieving at least one candidate neural network model from a model storage. 
     
     
         32 . The system of  claim 21 , wherein generating the plurality of candidate neural network models comprises overfitting at least one of the plurality of candidate neural network models to the input model. 
     
     
         33 . The system of  claim 21 , wherein tuning the plurality of candidate neural network models comprises:
 training at least one of the plurality of candidate neural network models; and   terminating the training when one or more training conditions are satisfied.   
     
     
         34 . The system of  claim 33 , wherein the one or more training conditions comprise at least one of a run time, a number of epochs, or an accuracy score. 
     
     
         35 . The system of  claim 33 , wherein training the at least one of the plurality of candidate neural network models comprises:
 sending a first command to a first development instance to train a first candidate neural network model; and   sending a second command to a second development instance different from the first development instance to train a second candidate neural network model.   
     
     
         36 . The system of  claim 21 , wherein the one or more model selection criteria comprise a selection criterium associated with an accuracy score of a candidate neural network model with respect to the input model features. 
     
     
         37 . The system of  claim 21 , wherein the one or more model selection criteria comprise a selection criterium associated with a model run time of a candidate neural network model. 
     
     
         38 . The system of  claim 21 , wherein generating the plurality of candidate neural network models comprises:
 sending a first command to a first development instance to generate a first candidate neural network model; and   sending a second command to a second development instance different from the first development instance to generate a second candidate neural network model.   
     
     
         39 . A method for transforming an input model into a neural network model, the method comprising:
 obtaining dataset model output by applying an input model to an input dataset, the input model having a plurality of input model features;   generating a plurality of candidate neural network models based on the input model features;   tuning the plurality of candidate neural network models by adjusting at least one of a plurality of hidden layers, a plurality of inputs, or a type of layer during tuning such that at least one of the input model features is reproduced;   selecting a neural network model from the plurality of candidate neural network models based on one or more model selection criteria; and   returning the selected neural network model.   
     
     
         40 . A non-transitory computer-readable medium storing instructions that, when executed by at least one processor, are configured to cause the at least one processor to perform operations comprising:
 obtaining dataset model output by applying an input model to an input dataset, the input model having a plurality of input model features;   generating a plurality of candidate neural network models based on the input model features;   tuning the plurality of candidate neural network models by adjusting at least one of a plurality of hidden layers, a plurality of inputs, or a type of layer during tuning such that at least one of the input model features is reproduced;   selecting a neural network model from the plurality of candidate neural network models based on one or more model selection criteria, a model type of the selected neural network model being different from a model type of the input model; and   returning the selected neural network model.