@BellmanDevs

Developing Bellman toolbox for model-based reinforcement learning, based on TensorFlow and TensorFlow Agents

bellman.dev
Joined March 2021
  • Tweets4
  • Following24
  • Followers19
Bellman has been originally developed at @SecondmindAI by @vincentadam87, @jordigraumo, @felix_leibfried, John McLeod, @HrvojeStojic and Peter Vrancx. It has been recently open-sourced, under Apache 2.0 licence, #ReinforcementLearning fans are welcome, come and contribute!
7
12
Hello world! Bellman is a package for model-based #ReinforcementLearning in #Python, using @TensorFlow. Version 0.1.0 has just been released on @pypi! Check our GH repo github.com/Bellman-devs/bell… and website at bellman.dev/.

GitHub - Bellman-devs/bellman: Model-based reinforcement learning in TensorFlow

Model-based reinforcement learning in TensorFlow. Contribute to Bellman-devs/bellman development by creating an account on GitHub.

github.com
6
8
more replies
Bellman boasts modular design, its meant to be easy to compose model-based agents, easily swapping supervised learning methods for learning dynamics, trajectory optimisers or background planners.
1
2
2
Bellman has several implementations of state-of-the-art model-based algorithms: PETS, MBPO and METRPO. Stay tuned for replication experiments!
2
2
Load more