Concurrent Data Sampling and Feature Selection via Weak Learners
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning (ML) practices face inefficiencies in feature selection and data sampling, as these steps are typically performed separately, leading to suboptimal resource consumption and increased complexity in training and inferencing processes.
Innovation Solution
A concurrent approach that unifies feature selection and data sampling by assigning weights to data instances and features, using weak learners to iteratively select a subset for training, thereby reducing the size of the training corpus and accelerating the process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If data sampling and feature selection are performed as two separate steps, then each step can be optimized independently, but the overall training time and space complexity increase
Solution Approach 1:
The patent combines data sampling and feature selection into a single unified process that simultaneously selects both data instances and features. This merging eliminates the need for two separate processing steps, reducing overall training time while maintaining model accuracy through coordinated selection of informative data-feature pairs.
Solution Approach 2:
The patent performs preliminary selection of data instances and features together before the main training process begins. By pre-selecting the most informative combinations of data and features in advance, the system reduces the computational burden during actual training, thereby decreasing training time without compromising model performance.
2Measurement precision
If more data instances and features are collected, then model accuracy and performance improve, but training time and space complexity increase
Solution Approach 1:
The patent extracts and selects only the most informative subset of data instances and features from the complete dataset. By taking out only the essential components needed for accurate modeling, the system maintains high model accuracy while significantly reducing the space complexity and memory requirements for training.
Solution Approach 2:
The patent applies local quality by selecting specific high-value data instances and features rather than using all data uniformly. This selective approach focuses computational resources on the most informative local regions of the data space, achieving high model accuracy with reduced space complexity by ignoring redundant or low-quality data.
3Ease of manufacture
If data sampling and feature selection are performed separately, then the implementation is simpler, but the resource consumption and training complexity increase
Solution Approach 1:
The patent merges data sampling and feature selection into a single integrated operation that simultaneously optimizes both data instance selection and feature selection. This unified approach improves training efficiency by eliminating redundant processing steps while maintaining implementation feasibility through a cohesive algorithmic framework.
Data Source
AI summary
From many features and many multidimensional points, a computer generates exploratory training configurations. Each point contains a value for each of the features. Each exploratory training configuration identifies a random subset of the features and a random subset of the points. A performance score is generated for each of the exploratory training configurations. A feature weight is generated for each of the features that is based on the performance scores of the exploratory training configurations whose random subset of features contains the feature. A point weight is generated for each of the points that is based on the performance scores of the exploratory training configurations whose random subset of the many points contains the point. A machine learning model is trained using an optimized training corpus that consists of a subset of the many features based on feature weight and a subset of the many points based on point weight.