Decision Tree Training with Difference Subsets for Classification Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The training time and computational load required for decision trees in image classification tasks are significant, making them inefficient for real-time applications.
Innovation Solution
An iterative approach is adopted where an initial tree is trained on a subset of features, and subsequent trees are trained only at nodes where the initial tree incorrectly classifies samples, using additional features to improve classification accuracy without burdening the processor with unnecessary computations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If decision trees are trained using all features and all training samples, then classification accuracy is improved, but training time and computational load increase significantly
Solution Approach 1:
The patent segments the training process into multiple stages: an initial tree is trained on a subset of features, and then subsequent trees are trained only on specific nodes where classification errors occurred. This divides the full training task into manageable segments, reducing overall computational load while maintaining accuracy through targeted refinement.
Solution Approach 2:
Instead of uniformly improving the entire decision tree, the patent applies local quality by focusing subsequent training only on specific nodes where errors occurred. Each subsequent tree addresses local weaknesses in the initial tree, training on a subset of features relevant to those specific error nodes, thereby reducing global computational complexity.
2Measurement precision
If decision trees are trained using all features and all training samples, then classification accuracy is improved, but computational load increases significantly
Solution Approach 1:
The training process is segmented into an initial training phase and subsequent refinement phases. The initial tree is trained on a subset of features with lower computational cost, and only the necessary subsequent trees are trained on specific nodes where errors occurred, reducing overall computational load while maintaining accuracy.
Solution Approach 2:
The patent applies partial action by training subsequent trees only on the subset of features and samples necessary to correct specific classification errors, rather than retraining the entire tree on all features. This partial refinement approach reduces computational load while achieving the necessary accuracy improvement.
3Loss of time
If an initial tree is trained on a subset of features, then training time is reduced, but classification accuracy may be insufficient
Solution Approach 1:
The patent implements feedback by evaluating the initial tree's classification performance and using the results to determine which nodes require subsequent training. The feedback from classification errors guides the selection of features and samples for subsequent trees, ensuring that time investment is focused on areas where accuracy can be improved.
Solution Approach 2:
The initial tree training serves as a preliminary action that establishes a baseline model quickly using a subset of features. This preliminary model then guides subsequent training efforts, allowing the system to achieve sufficient accuracy with minimal additional computational resources by focusing only on error-prone nodes.
Data Source
AI summary
A method may include obtaining a plurality of training samples with a plurality of classifications that include a first classification and a second classification, training an initial tree with an initial set of training samples selected from the plurality of training samples using an initial set of feature values extracted from the set of training samples, and, in response to determining that the initial tree incorrectly classified the initial set of training samples at an output node of the initial tree, training a subsequent tree using a subsequent set of feature values extracted from a subsequent set of training samples.