e0 = 0.90
e1 = 0.10
decay_fun = sch_exp_decay
cos_ann = True
ann_cyc = 5
schedule = Schedule(t0,
t1,
e0,
e1,
decay_fun,
cosine_annealing=cos_ann,
annealing_cycles=ann_cyc)
# Policy
policy = EpsilonGreedyPolicy(schedule=schedule, value_function=Q)
# Reward Function
reward_fun = rf_info2d_pos
# Action Pre/Post-Processing Action
act_fun = act_disc2cont
# Agent
lr = 1e-4
gamma = 0.99
doubleQ = True # Run doubleQ-DQN sampling from Q_target and bootstraping from Q
rb = True
rb_max_size = 1e6
rb_batch_size = 64
tau = 0.1