Related Experiment Videos
Comments on "a parallel mixture of SVMs for very large scale problems"
Xiaomei Liu1, Lawrence O Hall, Kevin W Bowyer
1Department of Computer Science and Engineering, University of Notre Dame, South Bend, IN 46556, USA. xliu5@cse.nd.edu
Neural Computation
|May 29, 2004
Summary
A novel neural network approach using support vector machines (SVMs) achieved 90.72% accuracy. However, an ensemble of decision trees proved more accurate (94.75%) and computationally efficient for classification tasks.
Area of Science:
- Machine Learning
- Computer Science
Background:
- A novel neural network approach was proposed for combining support vector machines (SVMs) in 2002.
- This method offered computational efficiency for large datasets.
Purpose of the Study:
- To compare the accuracy of the neural network-SVM ensemble with alternative classification methods.
- To evaluate the performance of decision tree ensembles on the Forest Cover dataset.
Main Methods:
- Implemented a neural network ensemble of support vector machines (SVMs).
- Developed and tested an ensemble of decision trees.
- Evaluated classifier performance on the Forest Cover dataset using an independent test set.
Main Results:
- The neural network-SVM ensemble achieved 90.72% accuracy.
- A simple ensemble of decision trees achieved a higher accuracy of 94.75%.
- The decision tree ensemble was also found to be computationally efficient.
Conclusions:
- Ensembles of decision trees can outperform neural network-SVM hybrids in accuracy.
- Computational efficiency is a key consideration in classifier selection.
- Experimental comparison of diverse classifier types is crucial for identifying optimal methods.