Open alexnix opened 7 years ago
I think so
Although depending on the complexity of the task maybe conventional fixed rate neuroevolution with a fixed network, or reinforcement learning may work more robustly for many tasks
On Mon, May 15, 2017 at 6:55 AM Alexandru Niculae notifications@github.com wrote:
Can I use this library only to evolve NNs using NEAT algorithm? No dataset is required as the fitness will result by using the NNs in a simulation and evaluating how well thy perform in controlling a robot.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/hardmaru/backprop-neat-js/issues/1, or mute the thread https://github.com/notifications/unsubscribe-auth/AGBoHvwSNogiPHK009PLpYP_lM3b1L38ks5r6FlUgaJpZM4NbK0F .
Can I use this library only to evolve NNs using NEAT algorithm? No dataset is required as the fitness will result by using the NNs in a simulation and evaluating how well thy perform in controlling a robot.