Bayesian Learning for Neural Networks
Bayesian Learning for Neural Networks
Sparse and Semi-supervised Visual Mapping with the S^3GP
CVPR '06 Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition - Volume 1
Computers and Electronics in Agriculture
Incorporating prior model into Gaussian processes regression for WEDM process modeling
Expert Systems with Applications: An International Journal
| Hi-index | 0.00 |
Taking into account the scarcity of feasible fertilizing rate feedback from on-board sensors and the limited computation power of controller mounted on the variable-rate granular fertilizer applicator, optimum control index chart is a good option for controller design to achieve accurate variable-rate fertilizing control. The index chart contains a list of optimum control parameters optimized to meet combined objectives: fertilizing accuracy, energy saving and fertilizing consistency. To generate such list, the probabilistic meta-model based on Gaussian Processes (GP) is firstly utilized to identify the variable-rate fertilizing process with the indoor experimental data. Consequently, the meta-model-based optimization process is presented with the given fertilizing rate, previous opening length and its adjusting direction. The optimal control parameters chart is obtained by the iterative multi-objectives optimization based on Genetic Algorithm (GA). Thus, the fertilizing prescription can be converted to the controller's actions by searching the optimum control parameters chart. The well-trained GP models predict the fertilizing rate and the fertilizing coefficient of variation with limiting mean relative error to 0.014 and 0.089, respectively. Finally, by considering four main error sources, the verified fertilizing model improves the average fertilizing rate error to no more than 5% at given management zone scale in field test.