Params Item Value loss_fn MSE grad_clip_flg None num_frames 1,000,000(1M) train_interval 100 nb_train_steps 50 learning_start After 50,000 steps memory_size 100,000 batch_size 100 gamma 0.99 L2_reg 0.5 soft_update_tau 0.01 optimiser(Actor) Adam(lr=1e-4) optimiser(Critic) Adam(lr=1e-3) Scores baselines result https://github.com/openai/baselines-results/blob/master/param-noise/mujoco.md My result Without Target Line With Target Line(Targets are derived from baselines website)
