New http://keras.io RL code example: applying Actor-Critic to the pole balancing problem, by @NandanApoorv
https://keras.io/examples/rl/actor_critic_cartpole/ …pic.twitter.com/h34zq9DzJ1
-
Very cool! This is also what keeps quadcopter drones in the air (which is how I'm familiar with the problem). Reinforcement learning is probably the least efficient way to solve this. A PID controller would suffice.
I'm reminded of Boston Dynamics' approach to robotics. While many people were trying to use RL to teach quadrupedal robots to walk, BD went for a ML-free approach based entirely on control theory. And got much better results, before anyone else.
-
-
Precisely. The mathematics behind this particular problem are well understood...which means that a) there is a direct and simple solution to it via symbolic computation and b) this is a demonstration of the computational equivalence of symbolic and neural methods.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.