Measurement Data Phase Assignment Using Reference Points for ML Generalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models used for evaluating vehicle or robot surroundings tend to overfit when trained with limited data, lacking the ability to generalize effectively to unseen data, especially in automated driving applications.
Innovation Solution
A method is introduced to divide measurement data records into distinct training phases by assigning them to reference points, using a sequence such as a Sobol sequence, ensuring similar data records are grouped together and optimizing and testing phases are separated to enhance generalization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a machine learning model is trained with a finite set of training examples to achieve accurate results, then the model can deliver accurate results for training examples, but it lacks the ability to generalize effectively to unseen data due to overfitting
Solution Approach 1:
The patent segments the training process into distinct phases by dividing the training data into multiple subsets, each associated with different reference points. This segmentation allows the model to learn from diverse data distributions without overfitting to any single subset, thereby improving generalization while maintaining training accuracy.
Solution Approach 2:
The patent applies local quality by associating different subsets of training data with specific reference points in the feature space. Each reference point represents a local region with specific characteristics, allowing the model to learn locally optimized features while maintaining global generalization capability through the diversity of reference points.
2Ease of manufacture
If measurement data records are randomly assigned to training phases, then the division process is simple, but similar data records may be assigned to different phases causing inconsistent training results
Solution Approach 1:
The patent introduces reference points as intermediaries between the training data and the phase assignment process. These reference points serve as mediators that systematically group similar data records together based on their proximity in the feature space, ensuring consistent assignment of similar records to the same phase while maintaining a structured and reproducible division process.
3Productivity
If all measurement data records are used in the optimization phase, then the model can be trained comprehensively, but the model may overfit and fail to generalize when tested on the same data
Solution Approach 1:
The patent segments the training data into multiple subsets associated with different reference points, and further divides these subsets across optimization and test phases. This ensures that the optimization phase uses comprehensive data for training while the test phase uses distinct subsets, preventing overfitting and ensuring reliable generalization performance.
Solution Approach 2:
The patent changes the parameter of data subset assignment by associating different reference points with different phases. This parameter change ensures that similar data records (those associated with the same reference point) are consistently assigned to the same phase, maintaining training comprehensiveness while enabling reliable generalization testing.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Method (100) for dividing a predetermined set of measurement data records (2) for training a machine learning model (1) into various predetermined phases (1a, 1b, 1c) of this training, wherein each measurement data record (2) contains values of one or more measured variables, comprising the steps: • a sequence of reference points (4) that cover a space (3) of the measurement data records (2) and do not coincide with measurement data records (2) is determined (110); • for one or more measurement data records (2) from the predetermined set of measurement data records (2), a predetermined distance measure (5) is used to determine (120) which reference point (4) this measurement data record (2) is closest to, and the measurement data record (2) is assigned (130) to this reference point (4); • the reference points (4) are divided (140) into the predetermined phases (1a, 1b, 1c) of the training, so that one or more reference points (4) are assigned to each phase (1a, 1b, 1c) of the training;• the measurement data records (2) assigned to this reference point (4) are also assigned (150) to the phase (1a, 1b, 1c) of the training to which each reference point (4) is assigned;