Automatic Slice Discovery and Tuning for Autonomous Vehicle Data Mining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data mining approaches for autonomous vehicles are labor-intensive and inefficient, requiring manual identification and tuning of data slices to address model weaknesses, which hampers the performance of machine learning models.
Innovation Solution
Implementing automatic slice discovery and tuning using event-based and model performance-based methods to identify and generate tailored datasets for training ML models, utilizing trained ML models to detect failure events and anomalies, and adjust thresholds for improved precision and recall.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual identification and tuning of data slices is used, then data mining can be performed, but the process becomes labor-intensive and inefficient
Solution Approach 1:
The system enables automatic slice discovery and tuning through machine learning models that autonomously identify relevant data slices and optimize their parameters without requiring manual intervention. The model automatically performs tasks that were previously done manually, including slice identification, parameter selection, and performance optimization.
Solution Approach 2:
The patent replaces the manual mechanical process of slice identification and tuning with an automated machine learning-based system. The ML model substitutes human operators in analyzing data patterns, identifying relevant slices, and tuning parameters, thereby eliminating labor-intensive manual operations while improving efficiency.
2Reliability
If manual slice tuning is performed, then data slices can be identified, but the process hampers the performance of machine learning models
Solution Approach 1:
The system implements a feedback mechanism where the machine learning model evaluates the performance of identified data slices and automatically adjusts slice parameters based on performance metrics. This closed-loop feedback ensures that slices are continuously optimized to improve ML model performance while maintaining efficient data mining operations.
Solution Approach 2:
The patent introduces dynamic adjustment of slice parameters based on real-time performance evaluation. Instead of static manual tuning, the system dynamically adapts slice definitions and parameters in response to changing data patterns and model performance requirements, enabling both high reliability and productivity.
3Productivity
If automatic slice discovery is implemented, then data mining efficiency improves, but system complexity increases
Solution Approach 1:
The patent employs a universal machine learning framework that handles multiple data slicing tasks through a single automated system. The ML model serves multiple functions including slice identification, parameter tuning, and performance evaluation, thereby managing system complexity through consolidation rather than proliferation of separate components.
Data Source
AI summary
Disclosed are embodiments for facilitating automatic slice discovery and slice tuning for data mining in autonomous systems. In some aspects, an embodiment includes providing, by a processing device hosting a slice discovery machine learning (ML) model, input data to the slice discovery ML model, the input data corresponding to performance data of an autonomous vehicle (AV); identifying, by the slice discovery ML model, attributes of the AV and corresponding thresholds for the attributes that define a slice comprising a collection of data sharing common characteristics; and providing, by the slice discovery ML model, the attributes and the corresponding thresholds defining the slice to a slice miner to mine training data corresponding to the slice for a tailored dataset.