Tools Used: Python, PyTorch, OpenAI Gym

Topics: Reinforcement Learning, Deep Learning, Neural Networks, Machine Learning


Co-Agent Networks is an actor-critic framework that utilizes some set of interactive nodes. We took this idea under investigation and developed multiple supervised and unsupervised settings that utilizes the nodes as actor-critic agents. The idea of Co-Agent systems is introduced by Philips S. Thomas, and is still being investigated.


We have developed a fully working framework that utilizes the nodes as actor-critic agents. We have completed multiple experiments and reported the results for the future researchers.

Read more about Co-MDP and Co-Agent Networks here:

To learn more about our implementation and results check out my blog post titled Co-agent Networks, Neural Network that think.