def set_environment(state_size, action_size):
state_count = int(state_size)
action_count_b = int(action_size)
global env
global agent
env = SON_environment(seed=seed)
agent = QLearner(seed=seed)
def set_environment(state_size, action_size):
state_count = int(state_size)
action_count_b = int(action_size)
global env
global agent
env = SON_environment(random_state=seed,
state_size=state_count,
action_size=action_count_b)
agent = QLearner(seed=seed,
state_size=state_count,
action_size=action_count_b,
batch_size=batch_size)
########################################################################################
radio_frame = 20
seeds = np.arange(100).tolist()
seeds = [98]
start_time = time.time()
for seed in seeds:
random.seed(seed)
np.random.seed(seed)
env = radio_environment(seed=seed)
agent = QLearner(seed=seed)
run_agent_fpa(env)
# run_agent_tabular(env)
# run_agent_deep(env)
# run_agent_optimal(env)
########################################################################################
end_time = time.time()
filename = 'figures/timing_M={}.txt'.format(env.M_ULA)
file = open(filename, 'w')
duration = 1000. * (end_time - start_time)
print('Execution time: {:4f} ms.\n'.format(duration))
file.write('Execution time: {:4f} ms.\n'.format(duration))
plt.step(np.arange(len(actions)), actions, color='b', label='Actions')
plt.savefig('figures/actions_episode_{}_seed{}.pdf'.format(
episode_index, seed),
format="pdf")
# plt.show(block=True)
plt.close(fig)
return
########################################################################################
radio_frame = 10
seeds = np.arange(500).astype(int).tolist()
#seeds = [0] # for optimal
for seed in seeds:
print('Now running seed: {}'.format(seed))
random.seed(seed)
np.random.seed(seed)
env = radio_environment(seed=seed)
# run_agent_optimal(env)
agent = QLearner(seed=seed) # only for the deep
run_agent_deep(env)
# K.clear_session() # free up GPU memory
del agent
########################################################################################
from DQNLearningAgent import DQNLearningAgent as QLearner
seed = 3 # change in Top_File.m also
random.seed(seed)
np.random.seed(seed)
batch_size = 32
state_count = 3
action_count_b = 5
env = SON_environment(random_state=seed,
state_size=state_count,
action_size=action_count_b)
agent = QLearner(seed=seed,
state_size=state_count,
action_size=action_count_b,
batch_size=batch_size)
# This is the entry point to the simulation
def env_reset_wrapper():
global env
global agent
state = env.reset()
return state
def agent_get_exploration_rate_wrapper():
global env
global agent
return agent.exploration_rate
ax.set_ylabel(r'$Q$')
ax_sec.set_ylabel(r'$L$')
plt.legend([plot1, plot2], [r'Average $Q$', r'Average loss'],
bbox_to_anchor=(0.1, 0.0, 0.80, 1),
bbox_transform=fig.transFigure,
loc='lower center',
ncol=3,
mode="expand",
borderaxespad=0.)
plt.tight_layout()
plt.savefig('output.pdf', format='pdf')
plt.show()
plt.close(fig)
seeds = np.arange(1).tolist()
for seed in seeds:
env = radio_environment(random_state=seed)
agent = QLearner(random_state=seed)
start_time = time.time()
run_agent_q(env)
end_time = time.time()
print('Simulation took {:.2f} minutes.'.format(
(end_time - start_time) / 60.))
########################################################################################