I'm procrastinating on my multiagent RL project for class by thumbing through Needham's Visual Differential Geometry. My god, it's full of pictures
catastrophic forgetter
- 9 Followers
- 30 Following
- 9 Posts
full time software dev, part time CS grad student, studying ML
@thegradient @BenjaminHan yeah, the main issue I see is that the only way to implement this currently is by forking mastodon, which is a maintenance burden since you have to periodically update the patch to the latest release. The maintainer already rejected the idea of merging this functionality into the main codebase (https://github.com/mastodon/mastodon/issues/822), so barring some kind of plugin system the only way to implement it is via a fork
@ashishgaurav_13 OpenAI's "Spinning Up" website has a page with the proof. Haven't fully gone through it yet though:
https://spinningup.openai.com/en/latest/spinningup/extra_pg_proof1.html?highlight=extra%20material
I was originally aggravated why neither "Grokking Deep RL" nor the GAE paper justified the interchangeability of total reward versus go-to-reward in the policy gradient. Surely the leap can't be that involved! Then I learned.
@BenjaminHan I may have a look at this when I get some time. looks like the mathstodon fork is available on github, so there may not be much work to do provided that we trust their implementation!
https://github.com/christianp/mastodon/tree/mathstodon-3.5.3
pretty jealous of the LaTeX support on mathstodon. that's one thing making me think that I should abandon social media and make blogging great again
I suppose setting up my own mastodon instance with LaTeX support is another option, albeit a more time-consuming one