SVM Kernel Parameter Optimization for Face Recognition Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine learning methods, such as Support Vector Machines (SVM), face challenges in achieving high recognition accuracy when the sample set is small, leading to data overfitting and increased false positives and false negatives.
Innovation Solution
The method involves optimizing the ratio of sample number to support vector number in an SVM model by selecting appropriate kernel parameters and iteratively refining the feature matrix to ensure this ratio exceeds a threshold, thereby reducing false positives and improving recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional SVM methods are used with small sample sets, then the model can be trained quickly, but recognition accuracy decreases due to data overfitting and high false positive rates
Solution Approach 1:
The patent applies preliminary action by performing automated feature selection and removing irrelevant or redundant features before training the SVM model. This preprocessing step ensures that only the most discriminative features are used, which prevents overfitting on small sample sets and reduces false positives while maintaining recognition accuracy.
Solution Approach 2:
The patent changes parameters by dynamically adjusting the feature subset based on statistical analysis and relevance criteria. By selecting optimal feature combinations and removing redundant ones, the system adapts the feature space to match the limited sample size, thereby improving generalization and reducing false positives without sacrificing accuracy.
2Measurement precision
If more features are included in the SVM model, then recognition completeness improves, but false positives increase due to noise and redundancy
Solution Approach 1:
The patent applies taking out by systematically removing irrelevant, redundant, and noisy features from the feature set. Through automated feature selection and statistical analysis, the system extracts only the most relevant features that contribute to accurate recognition, thereby maintaining recognition completeness while eliminating sources of false positives.
Solution Approach 2:
The patent applies local quality by assigning different weights and selections to different features based on their individual relevance and discriminative power. Instead of treating all features uniformly, the system selectively retains features with high local quality (high relevance) and removes those with low local quality (redundant or noisy), optimizing the balance between completeness and false positive reduction.
Data Source
AI summary
A computer system selects a kernel and kernel parameters for a first Support Vector Machine (SVM) model, testing the SVM model on a feature matrix T of n feature vectors of length m to produce false positive (FP) data set and false negative (FN) data set by a computer processor, wherein n and m are integer numbers, copies the feature matrix T to produce a feature matrix T_best, and checking if a ratio (T_best sample number)/(SVM support vector number) is above a threshold for the SVM model on T_best. If the ratio is above the threshold, SVM predictions are performed using the SVM model on the feature matrix T_best. The first SVM model can be used classify the faces or the objects in the images. An image-product design can be created based on the faces or the objects in the images classified by the first SVM model using the feature matrix T_best.