Machine learning lifecycle management
Systems, methods, and computer program products for managing a lifecycle of a machine learning (ML) application from a consumer point of view are described herein. Execution of an intelligent scenario for training of the ML application is initiated. An integrator component generates a training pipeline. The training pipeline includes training logic associated with a defined workflow for the training. An application having an input dataset trains the ML application using the training pipeline. The integrator component determines training metrics associated with the trained ML application. The training metrics are indicators of a level of accuracy of the trained ML application. A centralized component provides the training metrics for characterization of the trained model.
1. A method for managing a lifecycle of a machine learning (ML) application from a consumer point of view, the method implemented by one or more data processors forming part of at least one computing device and comprising:
initiating execution of an intelligent scenario for training of the ML application;
generating, by an integrator component, a training pipeline, wherein the training pipeline comprises training logic associated with a defined workflow for the training;
training, with an input dataset comprising a view of application data, the ML application using the training pipeline;
exposing an inference pipeline by application services, wherein the inference pipeline defines logic for transformations associated with an inference generated by the ML application;
determining, by the integrator component, training metrics associated with the trained ML application, wherein the training metrics are indicators of a level of accuracy of the trained ML application; and
providing, by a centralized component, the training metrics for characterization of the trained ML application.
2. The method of claim 1 , further comprising:
triggering, by the centralized component, training via an application programming interface; and
providing, by the integrator component, a training status and a training identification associated with the training.
3. The method of claim 2 , further comprising determining a status of the training by:
requesting, by the centralized component, a status update using the training identification;
returning, by the integrator component, updated status associated with the requested status update; and
updating, by the centralized component, training status with the returned updated status.
4. The method of claim 1 , wherein the ML application is created using a side-by-side ML architecture and the training pipeline is generated by a data intelligence platform.
5. The method of claim 1 , wherein the ML application is created using an embedded ML architecture and the training pipeline is generated by a predictive analytics integrator (PAi).
6. The method of claim 5 , wherein the embedded ML architecture is an in-memory database system.
7. The method of claim 1 , further comprising receiving, by the training pipeline, the input data set for the training, wherein the ML application is trained using the input dataset.
8. A system for managing a lifecycle of a machine learning (ML) application from a consumer point of view, the system comprising:
one or more data processors; and
memory storing instructions stored on at least one of the one or more data processors, which when executed, result in operations comprising:
initiating execution of an intelligent scenario for training of the ML application;
generating, by an integrator component, a training pipeline, wherein the training pipeline comprises training, with an input dataset comprising a view of application data, the ML application using the training pipeline;
exposing an inference pipeline by applications services, wherein the inference pipeline defines logic for transformations associated with an inference generated by the ML application;
integrating an inference API having a signature defined during design time as part of the intelligent scenario;
determining, by the integrator component, training metrics associated with the trained ML application, wherein the training metrics are indicators of a level of accuracy of the trained ML application; and
providing, by a centralized component, the training metrics for characterization of the trained ML application.
9. The system of claim 8 , wherein the operations further comprise:
triggering, by the centralized component, training via an application programming interface; and
providing, by the integrator component, a training status and a training identification associated with the training.
10. The system of claim 9 , wherein the operations further comprise determining a status of the training by:
requesting, by the centralized component, a status update using the training identification;
returning, by the integrator component, updated status associated with the requested status update; and
updating, by the centralized component, training status with the returned updated status.
11. The system of claim 9 , wherein the ML application is created using a side-by-side ML architecture and the training pipeline is generated by a data intelligence platform.
12. The system of claim 9 , wherein the ML application is created using an embedded ML architecture and the training pipeline is generated by a predictive analytics integrator (PAi).
13. The system of claim 12 , wherein the embedded ML architecture is an in-memory database system.
14. The system of claim 9 , wherein the operations further comprise receiving, by the training pipeline, the input data set for the training, wherein the ML application is trained using the input dataset.
15. A non-transitory computer program product storing instructions which, when executed by at least one data processor forming part of at least one computing device, implement operations comprising:
initiating execution of an intelligent scenario for training of a machine learning (ML) application;
triggering, by a centralized component, training via an application programming interface;
generating, by an integrator component, a training pipeline, wherein the training pipeline comprises training logic associated with a defined workflow for the training;
exposing an inference pipeline by applications services, wherein the inference pipeline defines logic for transformations associated with an inference generated by the ML application;
training, by an application having an input dataset, the ML application using the training pipeline;
providing, by the integrator component, a training status and a training identification associated with the training;
determining, by the integrator component, training metrics associated with the trained ML application, wherein the training metrics are indicators of a level of accuracy of the trained ML application; and
providing, by the centralized component, the training metrics for characterization of the trained ML application.
16. The non-transitory computer program product of claim 15 , further comprising determining a status of the training by:
requesting, by the centralized component, a status update using the training identification;
returning, by the integrator component, updated status associated with the requested status update; and
updating, by the centralized component, training status with the returned updated status.
17. The non-transitory computer program product of claim 15 , wherein the ML application is created using a side-by-side ML architecture and the training pipeline is generated by a data intelligence platform.
18. The non-transitory computer program product of claim 15 , wherein the ML application is created using an embedded ML architecture and the training pipeline is generated by a predictive analytics integrator (PAi).
19. The non-transitory computer program product of claim 18 , wherein the embedded ML architecture is an in-memory database system.