A multi-agent reinforcement learning environment tailored to collaborative and competitive gambling.
The aim was to test fMRI sequence generation and experiment with neurological signal behavior as a response to different gambling environments and rules. Other models, such as the fMRI Transformer Timeseries Conditional GAN or Sequence-to-Sequence models, work with this project.
The original project was initiated with my work with Neuromatch and is independently being continued here. The files utilize the RayRLLib framework and DQN agent for the Reinforcement Learning component. They are currently being formatted and organized to be shared here and for publication.
Feel free to contact me at robme.l@tuta.io.