Weed out Weak Agents

In Pezzza’s Work YT video, he makes a simulation where the weak agents are weeded out and the strong reproduce, and the children’s weights a changed slightly to give more variance. I was wondering if this type of behavior is possible using ML-Agents. I would say I’m a beginner with ML-Agents and I want to make something similar in Unity, could someone point me in the right direction?

Thank you!

ML-agents has no easy way to effect NN weights. The video looks like using genetic algorithm what is kind of different thing. But I think RL