def run(self):
self.totalList = [[] for i in range(0, self.T)]
for t in range (1, self.T):
crap = 0
for i in self.fireflies:
for j in i:
if j < 0 or j > 1:
print "Bounds error"
for i in range(0, self.Nff):
self.fireflies[i] = self.total_convergence(i, self.fireflies, self.fitnesses)
for i in range(0, self.Nff):
self.fireflies[i] = self.random_movements(self.fireflies[i], t, self.T)
for i in range(0, self.Nff):
self.fireflies[i] = np.sort(self.fireflies[i])
self.responses[i] = self.response_model.angleSpectrum(self.fireflies[i])
self.fitnesses[i] = self.response_model.calculateFitness(functions.convertDb(self.responses[i]))
self.average_fitnesses_over_time[t] = np.mean(self.fitnesses)
return self.fireflies, self.responses, self.fitnesses
def run(self):
self.totalList = [[] for i in range(0, self.T)]
for t in range (1, self.T):
# for i in self.fireflies:
# for j in i:
# if j < 0 or j > 1:
# print "Bounds error"
for i in range(0, self.Nff):
self.fireflies[i] = self.total_convergence(i, self.fireflies, self.fitnesses)
for i in range(0, self.Nff):
self.separations[i] = self.spacing_model.calculateSpacings(self.fireflies[i])
self.fireflies[i] = self.random_movements(self.fireflies[i], self.separations[i], t, self.T)
self.fireflies[i] = np.sort(self.fireflies[i])
self.responses[i] = self.response_model.angleSpectrum(self.fireflies[i])
self.fitnesses[i] = self.response_model.calculateFitness(functions.convertDb(self.responses[i]))
self.separations[i] = self.spacing_model.calculateSpacings(self.fireflies[i])
self.goodSeparationCount[i] = self.spacing_model.checkSpacings(self.separations[i])
self.averageGoodSeparationOverTime[t] = np.mean(self.goodSeparationCount)
self.averageFitnessOverTime[t] = np.mean(self.fitnesses)
self.averagePositionsOverTime[t] = np.mean(self.fireflies, axis=0)
return self.fireflies, self.responses, self.fitnesses, self.goodSeparationCount, self.averageGoodSeparationOverTime, self.averageFitnessOverTime, self.averagePositionsOverTime
def checkRandomUniformSampling(self):
fireflies = np.array([np.random.rand(self.N) for i in range (0, self.SAMPLES)])
count, bins = np.histogram(fireflies, bins=128, range=[0, 1], density=True)
plt.plot(bins[:-1], count)
plt.xlim([0,1])
plt.show()
responses = np.empty([self.SAMPLES,self.N_phi])
fitnesses = np.empty(self.SAMPLES)
for index in range (0, self.SAMPLES):
firefly = fireflies[index]
responses[index] = self.response_model.angleSpectrum(firefly)
fitnesses[index] = self.response_model.calculateFitness(functions.convertDb(responses[index]))
averageResponse = np.mean(responses, axis=0)
variance = np.var(responses, axis=0)
modeledAverage = self.generateModeledResponse()
modeledVariance = self.generateModeledVariance()
meanFitness = np.mean(fitnesses)
mse = ((averageResponse - modeledAverage) ** 2).mean(axis=None)
DbAverageResponse = functions.convertDb(np.mean(responses, axis=0))
DbModeledAverage = functions.convertDb(self.generateModeledResponse())
fig = plt.figure()
angleSpectrum = [(i*1.0/self.N_phi*180) for i in range (0, self.N_phi)]
plt.plot(angleSpectrum, DbAverageResponse, label="Simulation")
plt.plot(angleSpectrum, DbModeledAverage, label="Analytical Model")
plt.plot(angleSpectrum, self.response_model.getRt(), label="Desired")
plt.title("Mean Response Generated from\nRandom Uniform Sampling for Antenna Element Positions")
plt.legend()
plt.ylabel("Response (dB)")
plt.xlabel("Angle (degrees)")
plt.ylim([-65,5])
plt.show()
plt.plot(variance, label="Simulation")
plt.plot(modeledVariance, label="Analytical Model")
plt.legend()
plt.show()
return mse, DbAverageResponse
def generate_initials(self):
rand.seed()
self.fireflies = [np.sort(np.random.rand(self.N)) for i in range (0, self.Nff)]
self.responses = [self.response_model.angleSpectrum(self.fireflies[j]) for j in range (0, self.Nff)]
self.response_constraint = self.response_model.getRt()
self.fitnesses = [self.response_model.calculateFitness(functions.convertDb(self.responses[k])) for k in range (0, self.Nff)]
self.average_fitnesses_over_time = [0 for l in range (0, self.T)]
self.average_fitnesses_over_time[0] = np.mean(self.fitnesses)
def sampleFromCDF(self):
dfx = self.get_dataset()
cdf = self.generateCDF(dfx)
fireflies = np.array([np.array([self.sampleCDF(cdf) for j in range (0, self.N)]) for i in range (0, self.SAMPLES)])
count, bins = np.histogram(fireflies, bins=128, range=[0, 1], density=True)
plt.plot(bins[:-1], count)
plt.xlim([0,1])
plt.show()
responses = np.empty([self.SAMPLES,self.N_phi])
fitnesses = np.empty(self.SAMPLES)
for index in range (0, self.SAMPLES):
firefly = fireflies[index]
responses[index] = self.response_model.angleSpectrum(firefly)
fitnesses[index] = self.response_model.calculateFitness(functions.convertDb(responses[index]))
averageResponse = np.mean(responses, axis=0)
variance = np.var(responses, axis=0)
modeledAverage = self.generateModeledResponse()
modeledVariance = self.generateModeledVariance()
meanFitness = np.mean(fitnesses)
mse = ((averageResponse - modeledAverage) ** 2).mean(axis=None)
DbAverageResponse = functions.convertDb(np.mean(responses, axis=0))
DbModeledAverage = functions.convertDb(self.generateModeledResponse())
fig = plt.figure()
plt.plot(DbAverageResponse, label="Mean from CDF Sampled Antenna")
plt.plot(DbModeledAverage, label="Random Uniform Sampling")
plt.plot(self.response_model.getRt(), label="Desired")
plt.legend()
plt.ylim([-65,5])
plt.show()
plt.plot(variance, label="Simulation")
plt.plot(modeledVariance, label="Random Uniform Sampling")
plt.legend()
plt.show()
return mse, DbAverageResponse
def sampleUsingPercentileRanges(self):
dfx = self.get_dataset()
mins = np.percentile(dfx, 10.0, axis=0)
medians = np.percentile(dfx, 50.0, axis=0)
maxs = np.percentile(dfx, 90.0, axis=0)
fireflies = np.array([self.samplePercentileRange(mins, maxs, self.N) for i in range (0, self.SAMPLES)])
for i in range (0, 10):
x = fireflies[:,i]
count, bins = np.histogram(x, bins=128, range=[0, 1], density=True)
plt.plot(bins[:-1], count)
plt.xlim([0,1])
plt.show()
responses = np.empty([self.SAMPLES,self.N_phi])
fitnesses = np.empty(self.SAMPLES)
for index in range (0, self.SAMPLES):
firefly = fireflies[index]
responses[index] = self.response_model.angleSpectrum(firefly)
fitnesses[index] = self.response_model.calculateFitness(functions.convertDb(responses[index]))
averageResponse = functions.convertDb(np.mean(responses, axis=0))
variance= np.var(responses, axis=0)
medianResponse = functions.convertDb(self.response_model.angleSpectrum(medians))
meanFitness = np.mean(fitnesses)
fig = plt.figure()
plt.plot(averageResponse, label="Mean Sampling from Range 10th to 90th Percentile")
plt.plot(medianResponse, label="Response From Median")
plt.plot(self.response_model.getRt(), label="Desired")
plt.legend()
plt.ylim([-65,5])
plt.show()
plt.plot(variance, label="Simulation")
plt.legend()
plt.show()
return meanFitness, averageResponse
def analyze(self, fireflies, show=False):
responses = np.empty([self.SAMPLES, self.N_phi])
fitnesses = np.empty(self.SAMPLES)
for index in range(0, self.SAMPLES):
firefly = fireflies[index]
responses[index] = self.response_model.angleSpectrum(firefly)
fitnesses[index] = self.response_model.calculateFitness(
functions.convertDb(responses[index]))
averageResponse = functions.convertDb(np.mean(responses, axis=0))
variance = np.var(responses, axis=0)
meanFitness = np.mean(fitnesses)
if show == True:
for i in range(0, 10):
x = fireflies[:, i]
count, bins = np.histogram(x, bins=50, density=True)
plt.plot(bins[:-1], count)
plt.xlim([0, 1])
plt.title("Positions Generated from Linear Regression")
plt.show()
fig = plt.figure()
plt.plot(averageResponse, label="Mean from model")
# plt.plot(average, label="Theoretical")
plt.plot(self.response_model.getRt(), label="Desired")
plt.legend()
plt.ylim([-65, 5])
plt.show()
plt.plot(variance, label="Experimental")
plt.legend()
plt.show()
print meanFitness
return averageResponse, variance, meanFitness
def generate_initials(self):
# self.fireflies = [np.sort(np.random.rand(self.N)) for i in range (0, self.Nff)]
self.generateFireflies()
self.responses = [self.response_model.angleSpectrum(self.fireflies[j]) for j in range (0, self.Nff)]
self.response_constraint = self.response_model.getRt()
self.fitnesses = [self.response_model.calculateFitness(functions.convertDb(self.responses[k])) for k in range (0, self.Nff)]
self.averageFitnessOverTime = [0 for l in range (0, self.T)]
self.averageFitnessOverTime[0] = np.mean(self.fitnesses)
self.separations = [self.spacing_model.calculateSpacings(self.fireflies[i]) for i in range (0, self.Nff)]
self.goodSeparationCount = [self.spacing_model.checkSpacings(self.separations[i]) for i in range (0, self.Nff)]
self.averageGoodSeparationOverTime = np.empty(self.T)
self.averageGoodSeparationOverTime[0] = np.mean(self.goodSeparationCount)
self.averagePositionsOverTime = np.array([np.empty(self.N) for i in range (0, self.T)])
self.averagePositionsOverTime[0] = np.mean(self.fireflies, axis=0)
# filename = "data/2018-07-27_16.40/"
filename = "data/2018-08-02_16.32/"
dfx = pd.read_csv(filename + "x.csv", header=None)
dff = pd.read_csv(filename + "fitness.csv", header=None)
dfff = dff.values.flatten()
dropIndexs = []
fig = plt.figure()
count = 0
for i, val in enumerate(dfff):
antenna = dfx.iloc[i]
antenna = antenna.as_matrix()
response = model.angleSpectrum(antenna)
fitness = model.calculateFitness(functions.convertDb(response))
if fitness > 0.01:
dropIndexs.append(i)
count += 1
# plt.plot(functions.convertDb(response))
# plt.show()
# print antenna
print count
# print dropIndexs, len(dropIndexs)
dfx = dfx.drop(dropIndexs)
dfx.to_csv(filename + "xNew.csv", index=False, header=False)
dffNew = np.delete(dfff, dropIndexs)
np.savetxt(filename + "fNew.csv", dffNew)