Dynamic artificial intelligence / machine learning model update, or retrain and update, in digital processes at runtime

Patent metadata

Field	Value
Publication number	US-11822913-B2
Application number	US-201916722301-A
Country	US
Kind code	B2
Filing date	Dec 20, 2019
Priority date	Dec 20, 2019
Publication date	Nov 21, 2023
Grant date	Nov 21, 2023

Abstract

Dynamically updating, or retraining and updating, artificial intelligence (AI)/machine learning (ML) models in digital processes at runtime is disclosed. Production operation may not need to be stopped for AI/ML model update or retraining and update. The update steps and/or retraining steps for the AI/ML model may be included as part of the digital process. The AI/ML model update may be requested from internal logic (e.g., from the evaluation of a condition, by an that expression calls for the AI/ML model, etc.), external requests (e.g., from external triggers in a finite state machine (FSM), such as a file change, database data, a service call, etc.), or both. Automation of AI/ML model updates or retraining and updates may be provided, where the software reloads/reinitializes/re-instantiates with a retrained and/or updated AI/ML model after (and possibly immediately after) the AI/ML model becomes available.

First claim

The invention claimed is:

1. A computer-implemented method, comprising:
listening for an update request for an artificial intelligence (AI)/machine learning (ML) model, by a digital process executing on a computing system comprising a robotic process automation (RPA) workflow defining an execution order and a relationship between a set of activities including an activity that calls the AI/ML model is called by an expression of an activity of the set of activities; and
responsive to receiving the update request to update the AI/ML model, reinitializing or re-instantiating the digital process to call an updated version of the AI/ML model at runtime of the digital process by modifying the expression of the activity that calls the AI/ML model and listening for another update request, by the digital process executing on the computing system, wherein
the reinitializing of the digital process comprises resetting a state of one or more components of the digital process to an initial value, and
the re-instantiating of the digital process comprises creating the AI/ML model inside the digital process at runtime.

2. The computer-implemented method of claim 1 , further comprising:
listening for a retraining request for the AI/ML model, by the digital process executing on the computing system; and
responsive to receiving the retraining request to retrain the AI/ML model, initiating retraining of the AI/ML model during runtime of the digital process, by the digital process executing on the computing system.

3. The computer-implemented method of claim 2 , wherein the retraining of the AI/ML model occurs on one or more other computing systems different than the computing system executing the digital process.

4. The computer-implemented method of claim 2 , further comprising:
using a current version of the AI/ML model during the retraining, by the digital process running on the computing system, until the update request is received.

5. The computer-implemented method of claim 2 , further comprising:
comparing performance of the retrained AI/ML model to a performance threshold, against performance of a previous version of the AI/ML model, or both, by the digital process executing on the computing system; and
updating the AI/ML model responsive to the retrained AI/ML model exceeding the performance threshold, the performance of the previous version of the AI/ML model, or both, by the digital process executing on the computing system.

6. The computer-implemented method of claim 1 , further comprising:
automatically initiating retraining of the AI/ML model, by the digital process executing on the computing system, after a predetermined amount of training data is received, after a predetermined amount of time has elapsed since a last retraining, or both.

7. The computer-implemented method of claim 1 , wherein the digital process comprises a business process management (BPM) flowchart, a sequential flow, or a finite state machine (FSM).

8. The computer-implemented method of claim 1 , wherein the AI/ML model is embedded directly in the activity of the RPA workflow.

9. The computer-implemented method of claim 1 , wherein the digital process comprises an initialization state that loads the AI/ML model from storage or makes the AI/ML model callable by the digital process.

10. The computer-implemented method of claim 1 , further comprising:
receiving a request to run the AI/ML model, by the digital process running on the computing system;
executing the AI/ML model, or causing the AI/ML model to be executed, by the digital process running on the computing system; and
returning results of the execution of the AI/ML model, by the digital process running on the computing system.

11. A computer program comprising a digital process and embodied on a non-transitory computer-readable medium, the program configured to cause at least one processor to:
listen for a retraining request or an update request for an artificial intelligence (AI)/machine learning (ML) model comprising a robotic process automation (RPA) workflow defining an execution order and a relationship between a set of activities including an activity that calls the AI/ML model is called by an expression of an activity of the set of activities;
responsive to receiving the retraining request to retrain the AI/ML model, initiate retraining of the AI/ML model at runtime of the digital process; and
responsive to receiving the update request to update the AI/ML model, reinitialize or re-instantiate the digital process at runtime of the digital process by modifying the expression of the activity that calls the AI/ML model to call an updated version of the AI/ML model and listen for another retraining request or update request, wherein
the reinitializing of the digital process comprises resetting a state of one or more components of the digital process to an initial value, and
the re-instantiating of the digital process comprises creating the AI/ML model inside the digital process at runtime.

12. The computer program of claim 11 , wherein the program is further configured to cause the at least one processor to:
use a current version of the AI/ML model during the retraining until the update request is received.

13. The computer program of claim 11 , wherein the program is further configured to cause the at least one processor to:
compare performance of the retrained AI/ML model to a performance threshold, against performance of a previous version of the AI/ML model, or both; and
update the AI/ML model responsive to the retrained AI/ML model exceeding the performance threshold, the performance of the previous version of the AI/ML model, or both.

14. The computer program of claim 11 , wherein the program is further configured to cause the at least one processor to:
automatically initiate retraining of the AI/ML model after a predetermined amount of training data is received, after a predetermined amount of time has elapsed since a last retraining, or both.

15. The computer program of claim 11 , wherein the program is further configured to cause the at least one processor to:
receive a request to run the AI/ML model;
execute the AI/ML model or cause the AI/ML model to be executed; and
return results of the execution of the AI/ML model.

16. The computer program of claim 11 , wherein the AI/ML model is embedded directly in the activity of the RPA workflow.

17. A computing system, comprising:
memory storing computer program instructions comprising a digital process; and
at least one processor configured to execute the computer program instructions, the instructions configured to cause the at least one processor to:
listen for a retraining request for an artificial intelligence (AI)/machine learning (ML) model comprising a robotic process automation (RPA) workflow defining an execution order and a relationship between a set of activities including an activity that calls the AI/ML model is called by an expression of an activity of the set of activities,
responsive to receiving the retraining request to retrain the AI/ML model, initiate retraining of the AI/ML model at runtime of the digital process,
listen for an update request for the AI/ML model, and
responsive to receiving the update request to update the AI/ML model, reinitialize or re-instantiate the digital process at runtime of the digital process by modifying the expression of the activity that calls the AI/ML model to call an updated version of the AI/ML model and listen for another update request, by the digital process executing on the computing system, wherein

t

Assignees

• Uipath Inc

Inventors

• Oros Andrei Robert

Classifications

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06F8/65 Primary

  Updates (security arrangements therefor G06F21/57) · CPC title

• G06F8/71

  Version control (security arrangements therefor G06F21/57); Configuration management · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title