Particle Swarm Optimization (PSO) for feature selection enhanced with a local search algorithm

I have a project I will like for it to be implemented using python in jupyter notebook. It's a feature selection task using particle swarm optimization. Although, the particle swarm optimization involves initializing the particles using a different method in contrast to the randomize way. Also the pbest and gbest updating include some little modification which i will b glad to explain in details. Lastly, the final feature subset return will be refined using a local search algorithm called the Sequential One Point Flipping (SOPF) algorithm which goes through the final selected features sequentially and flips the state of each selected feature to see if a better accuracy can be found. (By flipping i mean unselecting a selected feature and then re-evaluate the feature subset to test if accuracy will be better than original)