This code is written in Python 2.7 and requires
- For the paper, we used a time step of 10 ms, which results in hundreds of time steps for typical tasks. When training a new task, we highly recommend that you start with a larger value to save time.
MIT
- Song, H. F., Yang, G. R., & Wang, X.-J. (2016) Reward-based training of recurrent neural networks for cognitive and value-based tasks.