def get_Tx_data(K, beta, SNR, C, dist, log_mse):
m = "Gaussian"
n = 2
r = 1
l = 1
c_type = "sym specific"
strong = 0.9
weak = 0
string = "AE_END"
if dist:
string = string + "_DIST"
if log_mse:
string = string + "_LOG"
snr = n * (10**(SNR / 10))
channel_name = get_channel_file_name(K, c_type, strong, weak, beta, C)
Channel = pickle.load(open(channel_name, "rb"))
A = Channel["A"]
H = convert_channel(A, K, SNR)
filename = get_filename(K, m, n, r, l, snr, c_type, strong, weak, beta, C,
"Saved_Models", string)
X1, VX1 = get_Tx_Output(K, m, n, r, l, SNR, H, A, filename)
file_name = get_filename(K, m, n, r, l, SNR, c_type, strong, weak, beta, C,
"Const_Data", string)
XY = {"X": X1, "Y": VX1}
pickle.dump(XY, open(file_name, "wb"))
def get_mse_ber_results(K, beta, SNR, C):
m = "Gaussian"
n = 2
c_type = "sym specific"
strong = 0.9
weak = 0
if m == "Gaussian":
r = 1
else:
r = m / n
l = 1
channel_name = get_channel_file_name(K, c_type, strong, weak, beta, C)
Channel = pickle.load(open(channel_name, "rb"))
A = Channel["A"]
H = convert_channel(A, K, SNR)
snr = 2 * (10 ** (SNR / 10))
model_file = get_filename(
K,
m,
n,
r,
l,
snr,
c_type,
strong,
weak,
beta,
C,
"Saved_Models",
"AE_TIMTIN_PRE",
)
if m == "Gaussian":
MSE_SNR = get_mse_ber(K, m, n, r, l, SNR, H, A, model_file)
else:
MSE_SNR, BER_SNR, SER_SNR = get_mse_ber(K, m, n, r, l, SNR, H, A, model_file)
file_name = get_filename(
K, m, n, 1, l, SNR, c_type, strong, weak, beta, C, "MSE_Data", "AE_TIMTIN_PRE"
)
if m == "Gaussian":
MSE = {"MSE": MSE_SNR}
pickle.dump(MSE, open(file_name, "wb"))
else:
MSE_BER_SER = {
"MSE": MSE_SNR,
"BER": BER_SNR,
"SER": SER_SNR,
}
pickle.dump(MSE_BER_SER, open(file_name, "wb"))
return MSE_SNR
def get_Tx_data(K, beta, SNR, C):
m = "Gaussian"
n = 2
r = 1
l = 1
c_type = "sym specific"
strong = 0.9
weak = 0
snr = n * (10 ** (SNR / 10))
channel_name = get_channel_file_name(K, c_type, strong, weak, beta, C)
Channel = pickle.load(open(channel_name, "rb"))
A = Channel["A"]
H = convert_channel(A, K, SNR)
filename = get_filename(
K,
m,
n,
r,
l,
snr,
c_type,
strong,
weak,
beta,
C,
"Saved_Models",
"AE_TIMTIN_PRE",
)
X1, VX1 = get_Tx_Output(K, m, n, r, l, SNR, H, A, filename)
file_name = get_filename(
K, m, n, r, l, SNR, c_type, strong, weak, beta, C, "Const_Data", "AE_TIMTIN_PRE"
)
XY = {"X": X1, "Y": VX1}
pickle.dump(XY, open(file_name, "wb"))
# K = 5
# beta = 0.5
# C = 0
# start_training_model(K, beta, 50, C, True, False)
# get_mse_ber_results(K, beta, [50])
# plot_noise_dist(K, beta, 50)
# MI_estimator(5, 0.9, [50])
# # # AE_multiprocess_save_model()
# for K in [3, 4, 5]:
# for beta in [0.5, 0.9]:
# AE_multiprocess(K, beta, [10, 30, 50])
# # AE_multiprocess_save_model()
def compute_data(K, beta, SNR, C):
m = "Gaussian"
n = 2
r = 1
c_type = "sym specific"
strong = 0.9
weak = 0
l = 1
channel_name = get_channel_file_name(K, c_type, strong, weak, beta, C)
Channel = pickle.load(open(channel_name, "rb"))
A = Channel["A"]
H = convert_channel(A, K, SNR)
snr = n * (10**(SNR / 10))
filename = get_filename(K, m, n, r, l, snr, c_type, strong, weak, beta, C,
"Saved_Models", "AE_PRE")
Network = pickle.load(open(filename, "rb"))
V = Network["V"]
U = Network["U"]
Rp = Network["R"]
if m == "Gaussian":
# MSE_SNR = maxsinr_comm(K, m, n, r, SNR, H, A)
MSE_SNR = maxSINR_ClosedForm(K, m, n, l, r, SNR, H, V, U, Rp)
# MSE_SNR = maxSINR_ClosedForm(K, m, n, l, r, SNR, H, A)
# MSE_SNR = maxsinr_comm_given(K, m, n, r, SNR, H, V, U, Rp)
else:
MSE_SNR, BER_SNR, SER_SNR = maxsinr_comm(K, m, n, r, SNR, H, A)
file_name = get_filename(K, m, n, r, l, SNR, c_type, strong, weak, beta, C,
"MSE_Data", "MaxSINR")
if m == "Gaussian":
MSE = {
"MSE": MSE_SNR,
}
pickle.dump(MSE, open(file_name, "wb"))
else:
MSE_BER_SER = {
"MSE": MSE_SNR,
"BER": BER_SNR,
"SER": SER_SNR,
}
pickle.dump(MSE_BER_SER, open(file_name, "wb"))
return MSE_SNR
def MI_estimator(K, beta, SNR, C):
m = "Gaussian"
n = 2
c_type = "sym specific"
strong = 0.9
weak = 0
if m == "Gaussian":
r = 1
else:
r = m / n
l = 1
channel_name = get_channel_file_name(K, c_type, strong, weak, beta, C)
Channel = pickle.load(open(channel_name, "rb"))
A = Channel["A"]
H = convert_channel(A, K, SNR)
snr = n * (10 ** (SNR / 10))
filename = get_filename(
K, m, n, r, l, snr, c_type, strong, weak, beta, C, "Saved_Models", "AE_PRE"
)
MI = get_mi_estimate(K, m, n, r, l, SNR, H, A, filename)
file_name = get_filename(
K, m, n, 1, l, SNR, c_type, strong, weak, beta, C, "MI_Data", "AE_PRE"
)
MI_ES = {
"MI": MI,
}
pickle.dump(MI_ES, open(file_name, "wb"))
return MI
# K = 5
# beta = 0.9
# # start_training_model(K, beta, [10], 350, 200, 100, True, False)
# get_mse_ber_results(K, beta, [40])
# plot_noise_dist(K, beta, 50)
# MI_estimator(5, 0.9, [50])
# # # AE_multiprocess_save_model()
# for K in [3, 4, 5]:
# for beta in [0.5, 0.9]:
# AE_multiprocess(K, beta, [10, 30, 50])
# # AE_multiprocess_save_model()
def MI_estimator(K, beta, SNR, C):
m = "Gaussian"
n = 2
c_type = "sym specific"
strong = 0.9
weak = 0
if m == "Gaussian":
r = 1
else:
r = m / n
l = 1
channel_name = get_channel_file_name(K, c_type, strong, weak, beta, C)
Channel = pickle.load(open(channel_name, "rb"))
A = Channel["A"]
H = convert_channel(A, K, SNR)
snr = n * (10 ** (SNR / 10))
filename = get_filename(
K,
m,
n,
r,
l,
snr,
c_type,
strong,
weak,
beta,
C,
"Saved_Models",
"AE_TIMTIN_PRE",
)
MI = get_mi_estimate(K, m, n, r, l, SNR, H, A, filename)
file_name = get_filename(
K, m, n, 1, l, SNR, c_type, strong, weak, beta, C, "MI_Data", "AE_TIMTIN_PRE"
)
MI_ES = {
"MI": MI,
}
pickle.dump(MI_ES, open(file_name, "wb"))
return MI
def MI_estimator(K, beta, SNR, C, dist, log_mse):
m = "Gaussian"
n = 2
c_type = "sym specific"
strong = 0.9
weak = 0
if m == "Gaussian":
r = 1
else:
r = m / n
l = 1
string = "AE_END"
if dist:
string = string + "_DIST"
if log_mse:
string = string + "_LOG"
channel_name = get_channel_file_name(K, c_type, strong, weak, beta, C)
Channel = pickle.load(open(channel_name, "rb"))
A = Channel["A"]
H = convert_channel(A, K, SNR)
snr = n * (10**(SNR / 10))
filename = get_filename(K, m, n, r, l, snr, c_type, strong, weak, beta, C,
"Saved_Models", string)
MI = get_mi_estimate(K, m, n, r, l, SNR, H, A, filename)
file_name = get_filename(K, m, n, 1, l, SNR, c_type, strong, weak, beta, C,
"Const_Data", string)
MI_ES = {
"MI": MI,
}
pickle.dump(MI_ES, open(file_name, "wb"))
return MI
def compute_mi_estimate(K, beta, SNR, C):
m = "Gaussian"
n = 2
r = 1
c_type = "sym specific"
strong = 0.9
weak = 0
l = 1
channel_name = get_channel_file_name(K, c_type, strong, weak, beta, C)
Channel = pickle.load(open(channel_name, "rb"))
A = Channel["A"]
H = convert_channel(A, K, SNR)
snr = n * (10**(SNR / 10))
filename = get_filename(K, m, n, r, l, snr, c_type, strong, weak, beta, C,
"Saved_Models", "AE_PRE")
Network = pickle.load(open(filename, "rb"))
V = Network["V"]
U = Network["U"]
Rp = Network["R"]
MI = get_mi_estimate(K, m, n, r, SNR, H, V, U, Rp)
file_name = get_filename(K, m, n, 1, l, SNR, c_type, strong, weak, beta, C,
"MI_Data", "MaxSINR")
MI_ES = {
"MI": MI,
}
pickle.dump(MI_ES, open(file_name, "wb"))
return MI
# plt.savefig("Figures/Final_MI.png")
# plt.show()
# Constellation
Beta = np.linspace(0.5, 1, 6)
string_list = [
"MaxSINR",
"AE_PRE",
"AE_TIMTIN_PRE",
]
C = 1
beta = 0.8
channel_name = get_channel_file_name(K, "sym specific", 0.9, 0, beta, C)
Channel = pickle.load(open(channel_name, "rb"))
A = Channel["A"]
fig, ax = plt.subplots(1, 3, sharey=True)
for i in range(3):
file2 = get_file(beta, C, 50, "Const_Data", string_list[i])
XY = pickle.load(open(file2, "rb"))
X = XY["X"]
Y = XY["Y"]
for k in range(K):
ax[i].scatter(X[k], Y[k][:, 0], label="User " + str(k))
def start_training_model(K, beta, SNR, C, save, check):
n1 = 500
n2 = 200
n3 = 200
m = "Gaussian"
n = 2
c_type = "sym specific"
strong = 0.9
weak = 0
channel_name = get_channel_file_name(K, c_type, strong, weak, beta, C)
Channel = pickle.load(open(channel_name, "rb"))
A = Channel["A"]
if m == "Gaussian":
r = 1
else:
r = m / n
l = 1
H = convert_channel(A, K, SNR)
snr = 2 * (10 ** (SNR / 10))
file_to_save = get_filename(
K,
m,
n,
r,
l,
snr,
c_type,
strong,
weak,
beta,
C,
"Saved_Models",
"AE_TIMTIN_PRE",
)
saved_file = get_filename(
K,
m,
n,
r,
l,
snr,
c_type,
strong,
weak,
beta,
C,
"Saved_Models",
"AE_TIMTIN_PRE",
)
training_model(
K,
file_to_save,
saved_file,
m,
n,
r,
l,
SNR,
H,
A,
n1,
n2,
n3,
save,
check,
)
def train_save_model(K, beta, SNR, C, dist, log_mse=False, new_model=False):
m = "Gaussian"
n = 2
c_type = "sym specific"
strong = 0.9
weak = 0
channel_name = get_channel_file_name(K, c_type, strong, weak, beta, C)
Channel = pickle.load(open(channel_name, "rb"))
A = Channel["A"]
if m == "Gaussian":
r = 1
else:
r = m / n
l = 1
n1 = 100
H = convert_channel(A, K, SNR)
snr = 2 * (10**(SNR / 10))
string = "AE_END"
if dist:
string = string + "_DIST"
if log_mse:
string = string + "_LOG"
file_to_save = get_filename(
K,
m,
n,
r,
l,
snr,
c_type,
strong,
weak,
beta,
C,
"Saved_Models",
string,
)
saved_file = get_filename(
K,
m,
n,
r,
l,
snr,
c_type,
strong,
weak,
beta,
C,
"Saved_Models",
string,
)
ae_train_save(
K,
m,
n,
r,
l,
SNR,
H,
A,
n1,
dist,
file_to_save,
saved_file,
False,
new_model,
True,
log_mse,
)
def compute_robust_data(K, beta, SNR, C, dist, log_mse):
m = "Gaussian"
n = 2
r = 1
c_type = "sym specific"
strong = 0.9
weak = 0
l = 1
channel_name = get_channel_file_name(K, c_type, strong, weak, beta, C)
Channel = pickle.load(open(channel_name, "rb"))
A = Channel["A"]
string = "AE_END"
if dist:
string = string + "_DIST"
if log_mse:
string = string + "_LOG"
H = convert_channel(A, K, SNR)
snr = 2 * (10**(SNR / 10))
model_file = get_filename(
K,
m,
n,
r,
l,
snr,
c_type,
strong,
weak,
beta,
C,
"Saved_Models",
string,
)
channel_name = get_perturbed_channel_file_name(K, c_type, strong, weak,
beta, C, 0)
Channel = pickle.load(open(channel_name, "rb"))
A1 = Channel["A"]
H1 = convert_channel(A1, K, SNR)
if m == "Gaussian":
MSE_SNR = get_mse_ber(K, m, n, r, l, SNR, H1, A1, model_file)
else:
MSE_SNR, BER_SNR, SER_SNR = get_mse_ber(K, m, n, r, l, SNR, H, A,
model_file)
file_name = get_filename(K, m, n, r, l, SNR, c_type, strong, weak, beta, C,
"Robust_MSE_Data", string)
if m == "Gaussian":
MSE = {
"MSE": MSE_SNR,
}
pickle.dump(MSE, open(file_name, "wb"))
else:
MSE_BER_SER = {
"MSE": MSE_SNR,
"BER": BER_SNR,
"SER": SER_SNR,
}
pickle.dump(MSE_BER_SER, open(file_name, "wb"))
return MSE_SNR