Validation of QT-Opt

[leinxx]

Hi Machina,

Great job in implementing the QT-Opt algorithm. This repo is the only full implementation of QT-Opt I found. Do you have validation results on the current implementation?

Thanks,

Lei

[rarilurelo]

We tried QT-Opt on Pendulum and reacher (ReacherBulletEnv-v0), and we confirmed to learn as these graphs show (compared with TRPO). Input information is not image but default input information of the envs in these experiments.