Policy gradient agent - no intention to add?

[jonilaserson]

I've noticed that the REINFORCE algorithm (aka policy gradient, without the Q function) is not listed in the list of agents, even the not-yet-implemented ones. I presume this was intentional? How come such a basic building block of RL was left out?

[RaphaelMeudec]

The list hasn't been updated recently, contributions on algorithm implementations are more than welcome!

[piyush01123]

Hi. I want to take this up. I'll submit a PR once I'm done.
From what I understand, I should create a class at rl/agents and write a test for it at tests/rl/agents. I'll run the tests as per the Contribution guide before submitting PR.

[tbmreza]

It has been months, is this issue closing anytime soon?

[random-user-x]

I have been working lately to add more algorithms to this software. I am currently busy with my thesis and examination. I will begin this work from mid May. Thank you :)

[tbmreza]

@mirraaj Great! If it's not too much, I would really really appreciate it if you make a blog post walking through the process creating it. Translating from equations to implementation has always been a mystical thing to me.

[random-user-x]

@tbmreza Sincere apologies for such a delay. I am back working for keras-rl. I will keep you updated. Look for new algorithms soon. :)

[hrik2001]

Hi! Any updates so far on this?