assert(pred_curr<=0.4)
#eps = [0.2, 0.12, 0.10, 0.08, 0.07, 0.06, 0.05, 0.01]
eps = [0.08, 0.07, 0.06, 0.05, 0.03, 0.01]
betas = data_pickle["betas"]
reward_types = [True]
res_r={}
for reward_type in reward_types:
res_b={}
for beta in betas:
thr=False
res_e={}
res_e[0]={"cnvg_val": trainer.test_curr_discriminator_batch(), "steps": 0}
for epsilon in eps:
print("t: "+str(n_authorized)+", snr: "+str(snr)+", eps: "+str(epsilon)+", beta: "+str(beta)+", reward: "+str(reward_type))
if thr:
res_e[epsilon]={"cnvg_val": trainer.test_curr_discriminator_batch(), "steps": 0}
else:
trainer.reset_generator(lr=1e-3, epsilon=epsilon, beta=beta, binary_reward=reward_type)
trainer.reflect_pre_train()
cnvg_val, steps = trainer.train_loop(steps=1000, sensitivity=sensitivity)
res_e[epsilon]={"cnvg_val": cnvg_val, "steps": steps}
if(cnvg_val < 0.01):
thr=True
res_b[beta]=res_e
res_r[reward_type]=res_b
res_t[snr]=res_r
data_pickle[n_authorized]=res_t
assert (pred_curr <= 0.4)
#eps = [0.2, 0.12, 0.10, 0.08, 0.07, 0.06, 0.05, 0.01]
eps = [0.05]
betas = [15000]
reward_types = [False]
res_r = {}
for reward_type in reward_types:
res_b = {}
for beta in betas:
thr = False
res_e = {}
res_e[0] = {
"cnvg_val": trainer.test_curr_discriminator_batch(),
"steps": 0
}
for epsilon in eps:
print("t: " + str(n_authorized) + ", snr: " + str(snr) +
", eps: " + str(epsilon) + ", beta: " + str(beta) +
", reward: " + str(reward_type))
if thr:
res_e[epsilon] = {
"cnvg_val":
trainer.test_curr_discriminator_batch(),
"steps": 0
}
continue
trainer.reset_generator(lr=1e-3,
epsilon=epsilon,