Boosted Classifier for Machine Learning with Malicious Noise
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine learning approaches are ineffective in generating accurate classifiers when faced with high rates of malicious noise, as they tend to produce erroneous results due to incorrectly labeled examples, limiting their accuracy and reliability.
Innovation Solution
The technique involves generating a boosted classifier through a boosting module that repeatedly calls a weak learner module to produce a plurality of weak classifiers, including those based on a majority vote of randomly-generated classifiers, which are selected and combined to improve accuracy despite malicious noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional machine learning approaches are used to process labeled examples, then the system can generate classifiers, but the accuracy deteriorates when malicious noise rates are high
Solution Approach 1:
The patent applies the principle of converting harm into benefit by using the presence of malicious noise to identify and exclude corrupted examples. The system processes labeled examples and identifies those with incorrect labels (malicious noise), then excludes them from the training set before generating the final classifier. This transforms the harmful effect of noise into a beneficial filtering mechanism that improves overall classifier accuracy.
2Reliability
If the system processes more labeled examples to improve accuracy, then the classifier performance improves, but the time required for processing increases
Solution Approach 1:
The patent applies preliminary action by performing example identification and exclusion before the main classification training process. The system first processes the labeled examples to identify and remove malicious noise, then uses the cleaned dataset for classifier generation. This preliminary purification step prevents time-consuming iterations and reprocessing, efficiently preparing the training data in advance.
Data Source
AI summary
A computationally efficient method and apparatus is disclosed for machine learning an unknown, n-dimensional γ-margin halfspace to accuracy 1−ε in the presence of malicious noise, when the noise rate is allowed to be as high asΘ(ɛγlog(1/γ)).A smooth boosting module repeatedly calls a weak learner module that generates candidate classifiers based on a majority vote of randomly-generated classifiers.