Simple Reinforcement Learning with Tensorflow Part 8: Asynchronous Actor-Critic Agents (A3C)
Arthur Juliani

Hi Arthur,

Thanks for the nice tutorials. Any particular reason you decided to implement workers as python threads and not real processes? Threads in python don’t really run in parallel, and job benefits only when there is a lot of I/O operations. Spinning workers as real processes would make them run on separate cores.

One more general question about your research. Few times you mentioned that you run theses networks on your laptop, and that you don’t really have computational resources for large scale nets and hyper parameter optimization. Considering the fast pace of advancement in this field, and resources that other people have on their disposal (mainly in big companies), how do you think that reflects on your research and ability to make scientific contribution?

Like what you read? Give Marko Simić a round of applause.

From a quick cheer to a standing ovation, clap to show how much you enjoyed this story.