To tune the parameters of Particle swarm optimization (PSO), there are two methods offline and online. In offline manner, the meta-optimization is used to tune the parameters of PSO by using another overlying optimizer. In the online manner, there are two techniques, Self-Adaptation, "consisting of adding some or all of the optimizers behavioural parameters to the search-space, thus making them subject to optimization along with the problem at hand". Another technique is Meta-Adaptation ," in which an overlaying optimizer is trying to tune the parameters of another optimizer in an online manner during the optimization of a problem."

"The concept of meta-optimization. A black-box optimizer is used in an offline manner as an overlaying meta-optimizer for finding good behavioural parameters of another optimization method, which in turn is used to optimize one or more actual problems."

In standard PSO the particles are initialized by using uniform random numbers and these particles are updated using update equations. the best solution is selected based on the best value of objective function.

In my work. I have two data sets, training and theoretical dataset and I need to initialize the particles by using training data instead of random numbers.

In this case, how can I tune the parameters of PSO using training and theoretical data set.

Also, I have a problem which is, I got the best cost in the initial step of PSO and in the initial step there are no parameters or update equations.

Is it possible to tune the parameters using Machine Learning method? How can I do this?



You must log in to answer this question.

Browse other questions tagged .