On-Device Anomaly Detection and User Feedback Loop
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for addressing anomalies in electronic device performance are curative and not preventive, fail to address reoccurrence, and do not account for user-specific experiences, lacking a mechanism for sharing remedies between users.
Innovation Solution
An on-device method that monitors parameters, identifies anomalies using an iterative re-trainable model, generates user-specific questions via a generative model, and shares user feedback to enhance user experience, with remote server collaboration for model retraining and anomaly prevention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional curative methods are used to address anomalies, then anomalies can be fixed after they occur, but the system cannot prevent anomalies from occurring or address reoccurrence
Solution Approach 1:
The system performs preliminary actions by continuously monitoring device parameters and using machine learning models to predict anomalies before they occur. The iterative re-trainable model analyzes historical data and current device state to proactively identify potential issues, allowing the system to take preventive measures rather than waiting for anomalies to manifest.
Solution Approach 2:
The system implements feedback mechanisms by collecting user responses to generated questions and using this feedback to retrain the machine learning models. User feedback loop enables the system to learn from actual user experiences and improve its anomaly detection accuracy over time, creating a continuous improvement cycle that enhances preventive capabilities.
2Measurement precision
If generic anomaly detection methods are used, then general anomalies can be detected, but user-specific experiences and reoccurrence patterns cannot be addressed
Solution Approach 1:
The system employs dynamic machine learning models that adapt and evolve over time. The iterative re-trainable model continuously updates its parameters based on new data and user feedback, allowing it to become more precise in detecting user-specific anomaly patterns without requiring complete model replacement. This dynamic adaptation improves detection accuracy while managing complexity through incremental learning.
Solution Approach 2:
The system performs self-service through automated model retraining using collected user feedback and device data. Rather than requiring manual intervention for model updates, the system automatically retrains its machine learning models with new information, improving its detection capabilities autonomously and reducing the complexity burden on users or operators.
3Adaptability or versatility
If individual user remedies are determined, then specific user issues can be addressed, but other users cannot benefit from shared learning
Solution Approach 1:
The system achieves universality by implementing a centralized feedback collection mechanism that aggregates data from multiple users. The machine learning models are trained on combined user feedback, enabling them to learn universal anomaly patterns that apply across different users while still maintaining the ability to handle user-specific variations. This multi-functional approach allows the system to serve individual users while benefiting from collective intelligence.
Solution Approach 2:
The system uses an intermediary feedback collection layer that mediates between individual user experiences and the central model training process. User-specific feedback is collected and processed through this intermediary layer, which then contributes to the overall model training. This intermediary mechanism ensures that individual user information is preserved and utilized while preventing data loss and enabling shared learning across the user base.
Data Source
Figure 1
Figure 2A~3
Figure 4
AI summary
Embodiments herein provide an on-device method for enhancing user experience in an electronic device. The method includes monitoring a plurality of parameters associated with an operation of the electronic device. The method includes identifying an anomaly associated with the electronic device based on the plurality of parameters associated with the operation of the electronic device and identifying a class of anomaly to which the anomaly associated with the electronic device belongs using a first on-device model. Further, the method includes presenting at least one question associated with the identified class of anomaly to user of the electronic device using on a second on-device model and receiving at least one user input for the at least one question. Furthermore, the method includes performing at least one action for enhancing the user experience based on the at least one user input for the at least one question.