Related Experiment Videos
Multicategory nets of single-layer perceptrons: complexity and sample-size issues
Sarunas Raudys1, Rimantas Kybartas, Edmundas Kazimieras Zavadskas
1Department of Informatics, Vilnius University, Vilnius, Lithuania. sarunas.raudys@mif.vu.lt
IEEE Transactions on Neural Networks
|March 11, 2010
Summary
This study proposes a new method for multicategory single-layer perceptrons (SLPs) to minimize classification error by refining cost functions and decision fusion. The approach enhances performance, especially in unbalanced datasets, outperforming traditional methods.
Area of Science:
- Machine Learning
- Pattern Recognition
- Artificial Intelligence
Background:
- Standard cost functions in multicategory single-layer perceptrons (SLPs) do not effectively minimize classification error rates.
- Existing methods struggle with unbalanced training datasets, necessitating improved classification strategies.
Purpose of the Study:
- To develop a novel approach for reducing classification error in multicategory SLPs.
- To introduce an unbalance correcting term for improved performance with imbalanced data.
- To compare different decision fusion methods and assess the effectiveness of SLP-based pairwise classifiers.
Main Methods:
- Rejection of traditional cost functions in favor of specialized SLP training and optimal stopping.
- Development and application of an unbalance correcting term for imbalanced datasets.
- Fusion of decisions using methods like Kulback-Leibler (K-L) distance and Wu-Lin-Weng (WLW).
- Utilizing colored noise injection to create pseudovalidation sets for finite sample problems.
Main Results:
- The proposed method effectively reduces classification error by optimizing SLP training and decision fusion.
- The unbalance correcting term significantly improves classification accuracy in unbalanced training set scenarios.
- Fusion methods based on K-L distance and WLW show similar performance for small sample sizes.
- SLP-based pairwise classifiers demonstrate comparable or superior performance to linear support vector classifiers in moderate dimensions.
- Colored noise injection proves effective for addressing finite sample issues in moderate-dimensional pattern recognition.
Conclusions:
- The refined SLP training, decision fusion, and unbalance correction offer a superior approach to minimize classification error.
- The proposed techniques provide robust performance, particularly in challenging unbalanced and moderate-dimensional pattern recognition tasks.
- The study highlights the potential of SLP-based pairwise classifiers as a competitive alternative to support vector machines.
Related Concept Videos
01:22Multi-input and Multi-variable systems
Cruise control systems in cars are designed as multi-input systems to maintain a driver's desired speed while compensating for external disturbances such as changes in terrain. The block diagram for a cruise control system typically includes two main inputs: the desired speed set by the driver and any external disturbances, such as the incline of the road. By adjusting the engine throttle, the system maintains the vehicle's speed as close to the desired value as possible.
In the absence of...
In the absence of...
01:25Neural Circuits
Neural circuits and neuronal pools are two of the main structures found in the nervous system. Neural circuits are networks of neurons that work together to carry out a specific task or process. They consist of interconnected neurons and glial cells, which provide structural and metabolic support.
Neuronal pools are collections of nerve cells with similar functions and interact through chemical and electrical signals. These pools include both interneurons (the central neural circuit nodes that...
Neuronal pools are collections of nerve cells with similar functions and interact through chemical and electrical signals. These pools include both interneurons (the central neural circuit nodes that...