simple-es VS evotorch

There have been many studies that combine RL and ES(evolutionary strategies), and combining these methods and multi-agent reinforcement learning is my current interest. As a one who has only studied RL and has no knowledge of ES, I have created a multi-agent evolutionary strategies project using pytorch, simple-es.

population based algos dont care about differentiability https://github.com/google/evojax https://github.com/RobertTLange/evosax https://github.com/nnaisense/evotorch https://github.com/uber-research/PyTorch-NEAT

Check out the release and try it now: https://github.com/nnaisense/evotorch/releases/tag/v0.3.0

EvoTorch (evotorch.ai) makes it straightforward to apply evolutionary reinforcement learning to the challenge. We've included setup help, training and visualisation scripts, a baseline controller trained through the provided script and help for submission to the competition. Simply head to the public GitHub to get started: https://github.com/nnaisense/evotorch-myosuite-starter

An Advanced Evolutionary library or algorithm has been a dream of scientists and AI/ML enthusiasts since the concept was introduced. This vision has come true thanks to the scientists at NNAISENSE, a Switzerland-based AI Enterprise. They created an open-source platform called EvoTorch. When operated in combination with Machine Learning, it can solve complex operational problems in a fraction of time, with lower costs, and at a larger scale. Evolutionary algorithms act as a step toward solving cascading problems that occur when the problem’s size and complexity increase. Evolutionary algorithms make the situations easier to handle the complexity without adding to the cost, they are also much easier to connect through GPUs and CPUs parallelly to ease up the calculation time and the complexity associated with it, that the only limit to your computational power becomes your budget. The evolutionary algorithms are built in the open framework EvoTorch.