Ordinal Mapping Neural Networks With Centroid-Free Loss
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing neural networks struggle to provide an efficient and optimized metric for ordinal classification, often requiring complex sampling and restrictive assumptions, and fail to effectively rank ordered classes without calculating hyperplane centroids.
Innovation Solution
The use of an ordinal mapping deep neural network trained with a centroid-free ordinal mapping loss function that minimizes distances between different classes and reduces overlap within classes, enabling direct estimation of ordered classes and reducing computational cost.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing neural networks use complex sampling and restrictive assumptions for ordinal classification, then measurement precision can be improved, but device complexity increases
Solution Approach 1:
The patent extracts and removes the complex sampling and restrictive assumptions from the neural network training process. By using a simplified loss function that directly optimizes for ordinal classification without requiring these complex components, the invention achieves accurate ordinal predictions while reducing network complexity.
Solution Approach 2:
Instead of using complex sampling and restrictive assumptions to achieve ordinal classification, the patent inverts the approach by using a simple loss function that directly learns the ordinal structure from the data without these complicating factors.
2Measurement precision
If existing neural networks calculate hyperplane centroids for ranking ordered classes, then measurement precision improves, but loss of time increases
Solution Approach 1:
The patent removes the hyperplane centroid calculation step from the ordinal classification process. By designing a loss function that directly optimizes ordinal rankings without requiring centroid computation, the invention achieves accurate class ranking while eliminating the associated computational time cost.
3Measurement precision
If existing neural networks use complex sampling methods, then measurement precision can be improved, but productivity decreases
Solution Approach 1:
The patent extracts and eliminates complex sampling methods from the training process. The simplified loss function works effectively with standard sampling approaches, achieving high ordinal classification accuracy while significantly improving training efficiency and productivity.
Data Source
AI summary
Training an ordinal mapping deep neural network (OMDNN) can include receiving multiple samples, each a computer-processable data structure corresponding to a real-world object and including a data element indicating one of n predefined classes to which each sample is linked. Each sample can be mapped by the OMDNN to sample points of a multidimensional space. The OMDNN can predicts the class of each sample based on an ordinal mapping. Parameters of the OMDNN can be iteratively adjusted in response to misclassifying one or more samples. Iteratively adjusting the parameters can be based on an expected loss determined by an ordinal mapping loss function that measures (a) distances between each sample point in the multidimensional space and each other sample point of the same class and (b) overlap between sample points of different classes.