ML Model Feedback for Network Failure Source Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models deployed in communication networks face challenges in identifying the source of network operation failures, leading to unnecessary re-training and potential degradation of model performance due to incorrect assumptions about failure causes.
Innovation Solution
A method is introduced where a device associated with a communication network performs tasks using a machine learning model, receives feedback data from an access node indicating the cause of network operation failures, and determines whether to re-train the model, update associated non-machine learning algorithm parameters, or refrain from such actions based on the feedback, allowing for targeted adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine learning models are continuously re-trained to improve performance, then model accuracy is improved, but time consumption and computational resources are increased
Solution Approach 1:
The patent implements a feedback mechanism where the device receives feedback data from the access node about network operation failures. This feedback information is used to determine whether re-training the machine learning model is necessary, allowing the system to learn from actual performance data and make informed decisions about model updates without unnecessary re-training cycles
Solution Approach 2:
The system performs preliminary assessment of failure causes before initiating re-training. By analyzing feedback data to determine whether the failure is device-related or network-related, the system prepares to re-train the model only when necessary, avoiding wasteful computational resources and time spent on unnecessary re-training operations
2Adaptability or versatility
If the device re-trains the machine learning model in response to any failure, then the model adapts to failures, but the model may be incorrectly assumed to be the failure source leading to unnecessary re-training
Solution Approach 1:
The patent uses feedback data from the access node that indicates the actual cause of network operation failures. This feedback mechanism allows the device to distinguish between device-related failures (where re-training is beneficial) and network-related failures (where re-training is unnecessary and may harm model stability). The feedback information is processed to make informed decisions about when to re-train the model
Solution Approach 2:
Instead of automatically re-training the model in response to any failure (the conventional approach), the patent inverts the logic by first analyzing the failure cause through feedback data and only initiating re-training when the failure is determined to be device-related. This inversion prevents unnecessary re-training and maintains model performance stability while still providing adaptability when needed
3Measurement precision
If feedback data is requested and processed to determine failure causes, then accurate failure identification is achieved, but additional communication overhead and processing complexity are introduced
Solution Approach 1:
The patent implements a feedback mechanism where the device requests and receives feedback data from the access node about network operation failures. This feedback information is processed to determine whether the failure is device-related or network-related, enabling accurate failure identification. The feedback data includes specific indicators that simplify the analysis process while maintaining high identification accuracy
Solution Approach 2:
The device autonomously processes the feedback data to determine failure causes without requiring external intervention. The self-service approach involves the device independently analyzing the feedback information, making decisions about whether to re-train the model, and managing its own optimization process, thereby reducing overall system complexity despite the additional feedback processing
Data Source
AI summary
Example embodiments may relate to controlling re-training of a machine learning (ML) model deployed at a device. A method may comprise: performing, by a device associated with a communication network, a task with a ML model to obtain an output, wherein the output is configured to be used for performance of a network operation of the communication network; receiving, from an access node of the communication network, feedback data indicative of a cause of a failure of the network operation; and determining, based on the feedback data, to perform at least one of the following: re-training the machine learning model for performing the task, updating at least one parameter of a non-machine learning algorithm associated with performance of the task with the machine learning model, refraining from re-training the machine learning model, or refraining from updating the at least one parameter of the non-machine learning algorithm.