원문정보
초록
영어
In order to increase the recognition rate of the CT image of benign or malignant pulmonary nodules, Support vector machine (SVM) was adopted to classify them. Meanwhile, Particle swarm optimization (PSO) algorithm is used to optimize parameters of the kernel function of SVM. Various optimization results were acquired through multiple methods such as consistent inertia weight, linear decreasing inertia weight, first increasing and then decreasing inertia weight and non-linear decreasing inertia weight. As was proved by experiments, recognition rates of these methods in training set were the same. The method with consistent inertia weight had a short optimizing time, but recognition rate in test set was low. The remaining methods had a long optimizing time while recognition effects were better in test set.
목차
1. Introduction
2. Method of Parameter Optimization
3. Standard Particle Swarm Optimization Algorithm
4. PSO with Inertia Weight
5. PSO Algorithms with Changing Inertia Weight
5.1. Inertia Weight Linear Decrease
5.2. Inertia WeightLlinear Differential Decrease
5.3. Inertia Weight First Increase then Decrease
5.4. Inertia Weight Nonlinear Decrease
6. Experiment and Simulation
7. Conclusion
References