Differential Evolution Evolved RBFNN based automated recognition of Traffic Sign Images
The neural network performances depend entirely upon the quality of learning it has obtained in the past. The conventional gradient-based approach may trap in the local minima, and it may be the cause of premature convergence. To overcome this difficulty, in this paper, a differential evolution that...
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | article |
| Lenguaje: | EN |
| Publicado: |
Elsevier
2021
|
| Materias: | |
| Acceso en línea: | https://doaj.org/article/892005ba52114c6a8a2784b37c7c5998 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | The neural network performances depend entirely upon the quality of learning it has obtained in the past. The conventional gradient-based approach may trap in the local minima, and it may be the cause of premature convergence. To overcome this difficulty, in this paper, a differential evolution that establishes optimal learning of Radial basis function neural network (RBFNN) has been proposed where different parameters of kernel function and connection weights have been evolved simultaneously. The benefits of evolutionary learning of radial basis function have been evaluated extensively against the gradient-based learning in the classification of benchmark XOR problem. The model integrates principal component analysis (PCA); evolutionary self-adaptive Radial basis function neural network and correlation difference-based decision modules. The PCA has been applied in the pre-processing stage to retrieve the core details available at the input side of the image. A self-adaptive form of radial basis function neural network is used to develop a classifier. For making the learning process more precise and faster, the adjustment parameters considered. One-to-one mapping is used in the learning process to obtain the correlation between input and network parameters. The high recognition efficiency and a high level of decision confidence are the benefits of the proposed hybrid model. With an increase in iteration, the learning error approaches zero value for Differential evolution Adaptive radial basis function (DEARBF). Further, this method is computationally efficient and has shown robustness against visibility limitation because of weather and the aging effect of traffic signboards, proving its applicability in practical life. |
|---|