Future of Reinforcement Learning (12.5)
Вставка
- Опубліковано 14 тра 2024
- Some thoughts on the future of reinforcement learning.
Code for This Video:
github.com/jeffheaton/app_dee...
~~~~~~~~~~~~~~~ COURSE MATERIAL ~~~~~~~~~~~~~~~
📖 Textbook - Coming soon
😸🐙 GitHub - github.com/jeffheaton/app_dee...
▶️ Play List - • 2023 PyTorch Version A...
🏫 WUSTL Course Site - sites.wustl.edu/jeffheaton/t8...
~~~~~~~~~~~~~~~ CONNECT ~~~~~~~~~~~~~~~
🖥️ Website: www.heatonresearch.com/
🐦 Twitter - / jeffheaton
😸🐙 GitHub - github.com/jeffheaton
📸 Instagram - / jeffheatondotcom
🦾 Discord: / discord
▶️ Subscribe: ua-cam.com/users/heatonresea...
~~~~~~~~~~~~~~ SUPPORT ME 🙏~~~~~~~~~~~~~~
🅿 Patreon - / jeffheaton
🙏 Other Ways to Support (some free) - www.heatonresearch.com/suppor...
~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#reinforcementlearning #gymnasium #PyTorch #DeepLearning #PyTorchTutorial #MachineLearning - Наука та технологія
The master of everything ML! used to use your c# ML libraries in the 90s!
RL is already used to solve incredible hard problems. One of the most important application is its use in organic chemistry for retrosynthesis. Another example is path planning in the milling process (5-axis milling machines with more then one head). Or optimizing of the bin packing into containers. Until now it is only hard to do because the models are comparable slow learners, the way to success is less straight forward and you need to find a way to simulate or model the relevant part of the real world environment where the problem to solve exists. A breakthrough in RL would be in my eyes more exiting than the current one of LLM's.
I do like RL a lot and can see so many applicable use cases. I have been using Ray, specifically RLlib as a framework that comes with some algos OOTB
Dr. Heaton would love to see more material from your channel on RL!
Some real world robots are trained in a simulator using reinforcement learning which is pretty cool. But I guess this the self-driving car example only with hands and/or legs.