3 ms·
I'd imagine part of any reinforcement learning approach would be randomly tweaking the physics for each run, to make it robust against small changes and errors.
by taneq 9y ago
I'd imagine part of any reinforcement learning approach would be randomly tweaking the physics for each run, to make it robust against small changes and errors. Handling the different games (assuming they're similar enough) should just be an extension of that.