Integrating heuristic optimization algorithms with SVRM toolbox

Hi

I am trying to model a nonlinear SVR for an experimental dataset containing 4 inputs and 1 output. The length of dataset is around 4500 out of which I have selected 3000 datasets randomly as training data. I'm using different nonparametric methods to model the data. For using Gaussian kernel SVR in the machine learning toolbox, I want to find the hyperparameters using optimization algorithms like PSO, with my own objective function. Can someone guide me how to integrate PSO with fitrsvm function, instead of default bayesopt.

TIA.