Multi-label Classification via Segmented Binary Predictors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Predictive models are limited in associating multiple labels with an input, as they typically assign a single label, making it difficult to determine complex patterns or characteristics, such as a data set being indicative of both vanilla and chocolate flavors simultaneously.
Innovation Solution
A system that trains multiple classifiers using a two-stage process, where individual classifiers generate intermediate outputs for each label, and a multi-label classifier combines these outputs to determine the applicability of multiple labels to a given input, allowing for the generation of multi-label predictive outputs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single predictive model is used to analyze data, then the model structure is simple and easy to implement, but the model cannot associate multiple labels with an input simultaneously
Solution Approach 1:
The patent divides the multi-label prediction task into multiple binary classification problems, where each classifier is responsible for predicting a specific label. This segmentation allows the system to handle multiple labels simultaneously while keeping each individual classifier relatively simple in structure.
Solution Approach 2:
The patent creates a universal classification framework where multiple classifiers work together to provide multi-functionality. Each classifier can be trained on specific training records and can handle different label types, making the overall system adaptable to various prediction tasks while maintaining a consistent structural approach.
2Measurement precision
If multiple classifiers are trained and combined, then the multi-label prediction accuracy improves, but the training data requirements and processing complexity increase
Solution Approach 1:
The training data is segmented into different subsets, with each subset dedicated to training a specific classifier for a particular label. This allows the system to distribute the training data requirements across multiple classifiers rather than requiring all classifiers to be trained on the complete dataset, thereby reducing the training burden for each individual model while maintaining overall prediction accuracy.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for multi-label models. One of the methods includes receiving training records, each training record having an input, a first output, and a second output. The method includes generating a first classifier using as input one of the inputs and using as output a corresponding one of the first outputs. The method includes generating a second classifier using as input one of the inputs and using as output a corresponding one of the second outputs. The method includes inputting the inputs into the first classifier and generating first predictive outputs. The method includes inputting the inputs into the second classifier and generating second predictive outputs. The method also includes generating a third classifier using as input the first output and the second output and using as output the first output and the second output of the corresponding training record.