Automated machine learning model selection

1. A computer-implemented method for identifying architectures of machine learning models meeting a user defined constraint, the computer-implemented method comprising:

receiving, by one or more processors, input associated with evaluating machine learning models from a user;

determining, by the one or more processors and using an inference time prediction component including a calibration component and an inference time analysis component, acceptable architectural templates to evaluate the machine learning models based on the input, wherein the determining of the acceptable architectural templates includes calibrating, with the calibration component, the performance of resource types associated with a pool of available resources by iterating through the permutations of the available resources measuring execution times for the resources;

determining, by the one or more processors, a first list of architectures and metrics based on a calculation of maximum neural network sizes of the acceptable architectural templates not exceeding the constraint, wherein the constraint is a monetary cost; and

sending, by the one or more processors, the first list of architectures and metrics to the user for selection.

2. The computer-implemented method of claim 1 , further comprising:

determining, by the one or more processors, a second list of architectures and metrics based on a calculation of maximum neural network sizes of the acceptable architectural templates not exceeding the constraint by more than a predetermined threshold; and

sending, by the one or more processors, the second list of architectures and metrics to the user for selection.

3. The computer-implemented method of claim 1 , wherein the input comprises a task, input types, output types, a dataset and a constraint.

4. The computer-implemented method of claim 1 , wherein the constraint further comprises a time limit for completing a task.

5. The computer-implemented method of claim 1 , wherein the metrics comprise task solution accuracy, inference time and inference monetary cost.

6. The computer-implemented method of claim 1 , wherein the metrics comprise a recommendation of an amount to reduce the constraint to accomplish a task.

7. The computer-implemented method of claim 1 , further comprising analyzing, with the interference time analysis component, a neural network based on determining an execution time cost for a given network by determining a time to execute individual blocks of the applicable neural network and then summing the execution times for the individual blocks of the neural network to obtain a performance time for the neural network.

8. A computer program product for identifying architectures of machine learning models meeting a user defined constraint, the computer program product comprising:

one or more non-transitory computer readable storage media and program instructions stored on the one or more non-transitory computer readable storage media, the program instructions comprising:

program instructions to receive input associated with evaluating machine learning models from a user;

program instructions to determine, using an inference time prediction component including a calibration component and an inference time analysis component, acceptable architectural templates to evaluate the machine learning models based on the input, wherein the determining of the acceptable architectural templates includes calibrating, with the calibration component, the performance of resource types associated with a pool of available resources by iterating through the permutations of the available resources measuring execution times for the resources;

program instructions to determine a first list of architectures and metrics based on a calculation of maximum neural network sizes of the acceptable architectural templates not exceeding the constraint, wherein the constraint is a monetary cost; and

program instructions to send the first list of architectures and metrics to the user for selection.

9. The computer program product of claim 8 , further comprising:

program instructions to determine a second list of architectures and metrics based on a calculation of maximum neural network sizes of the acceptable architectural templates not exceeding the constraint by more than a predetermined threshold; and

program instructions to send the second list of architectures and metrics to the user for selection.

10. The computer program product of claim 8 , wherein the input comprises a task, input types, output types, a dataset and a constraint.

11. The computer program product of claim 8 , wherein the constraint further comprises a time limit for completing a task.

12. The computer program product of claim 8 , wherein the metrics comprise task solution accuracy, inference time and inference monetary cost.

13. The computer program product of claim 8 , wherein the metrics comprise a recommendation of an amount to reduce the constraint to accomplish a task.

14. The computer program product of claim 8 , further comprising analyzing, with the interference time analysis component, a neural network based on determining an execution time cost for a given network by determining a time to execute individual blocks of the applicable neural network and then summing the execution times for the individual blocks of the neural network to obtain a performance time for the neural network.

15. A computer system for identifying architectures of machine learning models meeting a user defined constraint, the computer system comprising:

one or more computer processors;

one or more computer readable storage media; and

program instructions stored on the one or more computer readable storage media for execution by at least one of the one or more processors, the program instructions comprising:

program instructions to receive input associated with evaluating machine learning models from a user;

program instructions to determine, using an inference time prediction component including a calibration component and an inference time analysis component, acceptable architectural templates to evaluate the machine learning models based on the input, wherein the determining of the acceptable architectural templates includes calibrating, with the calibration component, the performance of resource types associated with a pool of available resources by iterating through the permutations of the available resources measuring execution times for the resources;

program instructions to determine a first list of architectures and metrics based on a calculation of maximum neural network sizes of the acceptable architectural templates not exceeding the constraint, wherein the constraint is a monetary cost; and

program instructions to send the first list of architectures and metrics to the user for selection.

16. The computer system of claim 15 , further comprising:

program instructions to determine a second list of architectures and metrics based on a calculation of maximum neural network sizes of the acceptable architectural templates not exceeding the constraint by more than a predetermined threshold; and

program instructions to send the second list of architectures and metrics to the user for selection.

17. The computer system of claim 15 , wherein the input comprises a task, input types, output types, a dataset and a constraint.

18. The computer system of claim 15 , wherein the constraint further comprises a time limit for completing a task.

19. The computer system of claim 15 , wherein the metrics comprise task solution accuracy, inference time, inference monetary cost and a recommendation of an amount to reduce the constraint to accomplish the task.

20. The computer system of claim 15 , further comprising analyzing, with the interference time analysis component, a neural network based on determining an execution time cost for a given network by determining a time to execute individual blocks of the applicable neural network and then summing the execution times for the individual blocks of the neural network to obtain a performance time for the neural network.