Machine Learning Model Classification Accuracy via Reliability Voting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing class classification methods using capsule networks often incorrectly classify unknown input data as known classes, leading to a desire for improved accuracy in class classification processing.
Innovation Solution
A method involving a machine learning model with multiple vector neuron layers, where input data is processed to generate classification output values and feature spectra, similarity degrees are calculated with known feature spectra, and reliability degrees are used to determine class votes, ultimately leading to accurate class determination results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a capsule network is used for class classification, then the classification processing can be performed, but unknown input data is erroneously classified as known classes
Solution Approach 1:
The patent introduces an intermediary verification mechanism between the capsule network's classification output and the final class determination. Specifically, it uses a feature spectrum obtained from a specific layer of the network, compares it with known feature spectra using similarity degree calculation, and uses this comparison as a mediator to verify whether the classification result is reliable before making the final class determination.
Solution Approach 2:
The patent implements a feedback mechanism where the similarity degree between the obtained feature spectrum and known feature spectra is used to provide feedback on the reliability of the classification result. This feedback loop allows the system to adjust its confidence in the classification and determine whether to accept the capsule network's output or reject it as unreliable, thereby preventing erroneous classifications of unknown data.
2Measurement precision
If multiple input data are generated from one target object and voted based on reliability degree, then class determination accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent segments the classification process into distinct stages: generating multiple input data variants, obtaining feature spectra from a specific layer, calculating similarity degrees against known feature spectra, determining reliability degrees, and performing weighted voting. This segmentation allows each component to be processed independently and systematically, making the complex overall process more manageable and implementable.
Solution Approach 2:
The patent performs preliminary actions by pre-obtaining and storing feature spectra from known classes before the actual classification task. During classification, instead of directly comparing raw input data with all possible classes, the system first extracts feature spectra from the capsule network's intermediate layers and compares these with the pre-computed known feature spectra, thereby simplifying the subsequent voting and determination process.
Data Source
AI summary
A method according to the present disclosure includes (a) generating N pieces of input data from one target object, (b) inputting the input data to a machine learning model and obtaining M classification output values, one determination class, and a feature spectrum, (c) obtaining a similarity degree between a known feature spectrum group and the feature spectrum for the input data, and obtaining a reliability degree with respect to the determination class as a function of the reliability degree, and (d) executing a vote for the determination class, based on the reliability degree with respect to the determination class, and determining a class determination result of the target object, based on a result of the vote.