Choosing Multiple Parameters for Support Vector Machines
Machine Learning
Prediction of chaotic time series based on the recurrent predictor neural network
IEEE Transactions on Signal Processing
An overview of statistical learning theory
IEEE Transactions on Neural Networks
Prediction of noisy chaotic time series using an optimal radial basis function neural network
IEEE Transactions on Neural Networks
Support vector machine with adaptive parameters in financial time series forecasting
IEEE Transactions on Neural Networks
Blind Image Deconvolution Through Support Vector Regression
IEEE Transactions on Neural Networks
| Hi-index | 0.00 |
The hyperparameters selection has a great affection on accuracy of support vector regression (SVR). In order to determine the hyperparameters of SVR, an adaptive chaotic cultural algorithm (ACCA) is employed for the optimal hyperparameters including kernel parameterss of Gaussian kernel function, regular constant γ and Ɛ in the Ɛ -insensitive loss function. Based on this, a learning algorithm with two-stage is constructed to realize the objective. Firstly, the initialization search spaces of hyperparameters are determined according to their influence on the performance of support vector regression. Secondly, optimal hyperparameters are selected using ACCA. ACCA adopt dual structure in cultural algorithm and adaptive chaotic mutation in evolution induction functions, and uses implicit knowledge extracted from evolution process to control mutation scale, which inducts individuals escaping from local best solutions. Taken the forecasting of gas concentration as example, experiment results indicate optimal hyperparameters can be obtained through above strategy.