Parallel SVM Ensemble for Large Dataset Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Support Vector Machines (SVMs) face performance degradation and increased computational resources requirements as dataset size grows, making them inefficient for large-scale classification tasks due to high computation and memory demands.
Innovation Solution
The method involves constructing an SVM ensemble by partitioning data into subsets, determining data clusters, identifying cross-classification matchups, and building multiple SVMs in parallel within each ensemble member, allowing for parallel processing and filtering to reduce data processing and memory usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a conventional SVM is used to handle large datasets, then classification accuracy can be maintained, but computation time and memory usage increase significantly
Solution Approach 1:
The patent divides the large dataset into multiple subsets and trains multiple SVMs in parallel on these subsets. Each SVM processes a portion of the data independently, and their results are combined through ensemble voting. This segmentation approach maintains classification accuracy while reducing the computation time and memory requirements for each individual SVM, thereby resolving the contradiction between maintaining accuracy and reducing computation time.
2Measurement precision
If a conventional SVM is used to handle large datasets, then classification accuracy can be maintained, but memory usage increases significantly
Solution Approach 1:
The patent segments the large dataset into multiple smaller subsets, each processed by a separate SVM. This reduces the memory footprint of each individual SVM while maintaining overall classification accuracy through ensemble combination. The memory usage is distributed across multiple parallel processes rather than concentrated in a single SVM, resolving the contradiction between maintaining accuracy and reducing memory usage.
Solution Approach 2:
The patent uses sampling techniques to create subsets that are representative of the full dataset but require less memory. By processing multiple partial subsets in parallel and combining results, the system achieves accurate classification without loading the entire large dataset into memory simultaneously, thus resolving the memory usage contradiction.
3Quantity of substance
If the size of the input dataset increases, then more data can be processed, but computational resources exceed a single processor's capacity
Solution Approach 1:
The patent divides the large dataset into multiple subsets that can be processed in parallel by multiple SVMs. This segmentation allows the system to handle larger datasets by distributing the computational load across multiple independent processing units, avoiding the need for a single processor to handle the entire dataset, thus resolving the contradiction between processing large datasets and managing computational resources.
Solution Approach 2:
The patent transitions from sequential processing on a single processor to parallel processing across multiple processors or computing nodes. By adding the dimension of parallelism, the system can process larger datasets without overloading a single processor, effectively resolving the resource capacity contradiction.
4Quantity of substance
If the size of the input dataset increases, then more data can be processed, but elapsed wall clock time becomes quite large
Solution Approach 1:
The patent segments the dataset into multiple subsets processed in parallel by multiple SVMs. This parallel processing significantly reduces the elapsed wall clock time compared to sequential processing of the entire dataset, while still enabling the system to handle large datasets. The contradiction between processing large datasets and reducing elapsed time is resolved through parallel execution of segmented tasks.
Data Source
AI summary
Systems, methods, and computer-readable media for building ensemble members of a Support Vector Machine (SVM) ensemble in parallel and executing processing in parallel on data allocated to each ensemble member are disclosed. The parallel construction and processing of data of each ensemble member allows a single large SVM calculation to be replaced with many smaller SVM calculations performed in parallel, and thus, may reduce the computational resources required to classify datasets.