Loading video...
Video Failed to Load
HydroGym: A Reinforcement Learning Platform for Fluid Dynamics Now published in Nature!! GitHub: Amazing collaboration with Christian Lagemann, S Mokbel, M Gondrum, M Rüttgers, Y Wang, P Suárez, L Paehler, D A Bezgin, A B Buhendwa, J L Callaham, S Ahnert, N Zolman, X Shao, J-Ch Loiseau, N A.... show more
92,125 views • 2 months ago •via X (Twitter)
23 Comments

@morgymcg Wow, incredible, congratulations Steve!

@morgymcg Thank you!! 🙏

Super excitied! Do you plan to release some datasets? I saw a folder for OfflineRL but it's still empty.

Yes in the works!

Best channel on ML and modeling.

This is the worst possible name for the awesomest physics engine i can imagine. also i wanna be a dolphin.

😂

🙌🔥. Can't wait to play around with the repo

The only best teacher I've ever seen💥💥

U my personal hero dude

With my regards Steve...

Congratulations!!!!

Turning RL into practical scientific simulation is seriously impressive

Some awesome work! Excited to see how this repo progresses

Breaks my mind how much still needs to be taught to agents

Huge win for RL in physics sims.

Amazing , HydroGym is exactly the kind of shared, physics-grounded RL infrastructure the field has needed.

@alfcnz @grok summarize the novelty and the setup in this paper

Cool 👍

does hydrogym expose solver instability in the reward, because agents will exploit it

Fluid dynamics meets RL, huge work

Hi. I would really like to hear you opinion on the ai discovery about the Navier Stokes equation and about what does it means for scientists.

Improved and extended this HydroGym paper using my new Ram Calculus thery. See for some details on Ram Calculus invented by me. I have similarly improved/extended over 2,300 other specific papers/patents published in the last 5 years, in are areas of Science, Math, and Engg. Contact me for details at [email protected].
