Vehicle Classification Error Handling Without Neural Network Retraining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Neural networks in vehicles often produce erroneous classification decisions, which are time-consuming and costly to correct through retraining, necessitating a more efficient solution.
Innovation Solution
A method, system, and computer-readable medium that automatically and systematically identify and correct neural network erroneous classification decisions without retraining the network, maintaining the network's state unchanged by using an analyze unit with auto-labeling capabilities and a fix unit to generate and apply signatures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If neural network retraining is performed to correct erroneous classification decisions, then classification accuracy is improved, but time consumption and cost increase significantly
Solution Approach 1:
The patent segments the correction process into two independent parts: (1) neural network classification for initial detection, and (2) rule-based post-processing for error correction. This segmentation allows the neural network to continue operating without retraining while a separate module handles corrections, thus improving accuracy without increasing retraining time.
Solution Approach 2:
The patent introduces an intermediary rule-based processing layer between the neural network output and final classification decisions. This intermediary layer validates and corrects neural network outputs using predefined rules, acting as a mediator that improves reliability without requiring the neural network itself to be retrained.
2Reliability
If neural network retraining is performed to correct erroneous classification decisions, then classification accuracy is improved, but software update complexity and cost increase
Solution Approach 1:
The patent separates the correction functionality into an independent rule-based module that does not require neural network retraining. This segmentation means that software updates only need to modify the rule-based module rather than retraining and redistributing the entire neural network, significantly reducing update complexity.
Solution Approach 2:
The patent creates a copy of the classification decision process in the form of rule-based validation logic. This copy operates in parallel to verify and correct neural network outputs, providing a lightweight alternative that avoids the complexity of neural network retraining and deployment.
3Reliability
If real-time correction of classification errors is implemented, then system reliability is improved, but processing complexity increases
Solution Approach 1:
The patent introduces an intermediary rule-based validation layer that processes neural network outputs in real-time. This intermediary uses simple, predetermined rules to verify and correct classifications, adding minimal processing complexity while significantly improving system reliability through immediate error correction.
Solution Approach 2:
The system performs self-correction by automatically validating and correcting its own outputs using the rule-based module. This self-service mechanism improves reliability in real-time without requiring external intervention or complex retraining processes, as the system corrects its own errors autonomously.
Data Source
AI summary
A method for real time management of detected issues, the method includes producing, by a classification unit having a neural network, a classification decision for sensed information obtained in an environment of a vehicle; generating, by one or more computing devices with auto-labeling capabilities, an automated ground truth labeling for the sensed information; detecting, by the one or more computing devices and based on a performance indication related to the automated ground truth labeling, an issue with respect to the classification decision; and responsive to the detecting, addressing the detected issue in a driving in the environment of the vehicle by a computer device associated with the vehicle, using a signature generated in association with at least the classification decision or with the detected issue. The neural network is in a same state in the producing of the classification decision, the detecting the issue, and the addressing the detected issue.