Robust Classifier Using Weighted Adder for Missing Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for handling datasets with missing or incomplete data, such as data imputation and building multiple classifiers, are either inaccurate or impractical due to the combinatorial explosion of feature subsets, necessitating a more robust approach for classification.
Innovation Solution
Implementing a system with low-dimensional classifiers associated with individual features, where a weighted adder adjusts weights based on the presence or absence of features in the data, producing a robust classification output, such as a hemodynamic instability score, without requiring complete data sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If data imputation is used to handle missing data, then implementation is simple, but classification accuracy deteriorates due to noise injection
Solution Approach 1:
The patent segments the classification task into multiple low-dimensional classifiers, each handling a specific feature subset. This allows the system to process incomplete data without imputation by distributing the classification burden across specialized classifiers, each trained on specific feature combinations.
Solution Approach 2:
The system dynamically changes parameters (weights) of classifiers based on the actual data present in each input. The weighted adder adjusts classifier contributions according to which features are actually measured, adapting to the specific pattern of missing data without requiring imputation.
2Measurement precision
If multiple classifiers on different feature subsets are built to handle missing data, then classification accuracy improves, but device complexity explodes combinatorially
Solution Approach 1:
Instead of creating classifiers for all possible feature subsets, the patent segments the problem into a manageable set of low-dimensional classifiers, each focusing on specific features. This segmentation avoids the combinatorial explosion while maintaining the ability to handle various missing data patterns.
Solution Approach 2:
The weighted adder serves multiple functions: it aggregates classifier outputs, dynamically adjusts weights based on missing features, and produces the final classification. This multi-functionality reduces the need for separate specialized components for each feature combination.
3Measurement precision
If complete datasets are required for accurate classification, then measurement precision improves, but adaptability to real-world incomplete data deteriorates
Solution Approach 1:
The system dynamically adjusts the weights of classifiers based on which features are present or missing in the input data. This dynamic adaptation allows the classifier to maintain accuracy whether data is complete or incomplete, automatically reconfiguring for the specific data pattern encountered.
Solution Approach 2:
The weights assigned to each classifier are changed based on the actual data present. When certain features are missing, the system changes the parameters (weights) of relevant classifiers to compensate, maintaining classification accuracy without requiring complete datasets.
Data Source
AI summary
Various embodiments described herein relate to methods and apparatus for robust classification. Many real-world datasets suffer from missing or incomplete data. By assigning weights to certain features of a dataset based on which feature(s) are missing or incomplete, embodiments of the prevention can provide robustness and resilience to missing data.