def solutionTest(func, ns, dataNum, lnLambdas=[], SavePath=""):
list = []
for i in ns:
listi = []
list.append(listi)
for lx in dataNum:
listx = []
listi.append(listx)
X, T = generateData(lx)
testX, testT = generateData(lx // 2)
if len(lnLambdas) == 0:
W = func(i, X, T)
print("%d;%d;None;%r" %
(i, lx, RSS(testT, testX, W) / len(testX)))
listx.append(W)
else:
for lnLambda in lnLambdas:
W = analyticalSolve(i, X, T, lnLambda)
listx.append(W)
print("%d;%d;%.0f;%r" %
(i, lx, lnLambda, RSS(testT, testX, W) / len(testX)))
for i in range(len(ns)):
for j in range(len(dataNum)):
rW = []
labels = []
for k in range(len(lnLambdas)):
rW.append(list[i][j][k])
labels.append("λ e^%.0f" % (lnLambdas[k]))
visualPoly(*rW,
*labels,
title="%s poly%d datanum%d" %
(func.__name__, ns[i], dataNum[j]),
savePath=SavePath)
def __init__(self):
print("Enter x range [A, B]\n")
x_A = float(input("A = "))
print('\n')
x_B = float(input("B = "))
print('\n')
step = float(input("Step = "))
print('\n')
print("Enter noise range [C, D]\n")
n_C = float(input("C = "))
print('\n')
n_D = float(input("D = "))
print('\n\n')
self.xValues = arange(x_A, x_B, step)
self.yValues = generateData(self.xValues, [n_C, n_D])
break
sum /= 2
W.reverse()
if isaverage:
sum /= count
return sum
if __name__ == '__main__':
dataNum = 10
n = 9
lr = 0.5
maxItrTimes = sys.maxsize
batch = dataNum
lnLambda = 0
X, T = generateData(dataNum)
e, resultW = gradientDescent(n,
X,
T,
lr,
lnLambada=lnLambda,
batch=batch,
maxItrTimes=maxItrTimes)
visualPoly(resultW,
"λ e^%.0f" % (lnLambda),
X=X,
T=T,
title="%s poly%d datanum%d lr%.3f batch%d" %
("gradientDescent", n, dataNum, lr, batch),
isShow=True,
lambada = math.e ** lnLambada
XX = mat([[x ** i for i in range(lenW)] for x in X])
vectorT = mat(T).T
A = XX.T * XX + lambada * numpy.eye(lenW) # 带惩罚项
B = XX.T * vectorT
W = mat(zeros((lenW, 1)))
r = B - A * W
p = r.copy()
num = 0
while num < MaxIterationNum:
num += 1
alpha = (r.T * r / (p.T * A * p))[0, 0]
W += alpha * p
lastr = r.copy()
r -= alpha * A * p
if (r.T * r)[0, 0] < limit:
break
beta = (r.T * r / (lastr.T * lastr))[0, 0]
p = r + beta * p
return W.T.tolist()[0], num
if __name__ == '__main__':
X, T = generateData(10)
W, num = conjugateGradient(9, X, T, lnLambada=None)
visualResultAndSampleAndTarget(W,X,T)
# plt.plot(X, T, 'r*', linewidth=2)
# plt.show()
from matplotlib import pyplot as plt
import numpy as np
from DataGenerator import generateData
from Settings import (DEFAULT_NUMBER_OF_CLASSES, DEFAULT_NUMBER_OF_FEATURES,
DEFAULT_NUMBER_OF_RECORDS_PER_CLASS,
DEFAULT_FEATURE_MEAN_RANGE, DEFAULT_RANDOM_NUMBER_SEED,
DEFAULT_TRUE_NUMBER_OF_CLASSES)
data, labels = generateData(DEFAULT_NUMBER_OF_CLASSES,
DEFAULT_NUMBER_OF_FEATURES,
DEFAULT_NUMBER_OF_RECORDS_PER_CLASS,
DEFAULT_FEATURE_MEAN_RANGE,
DEFAULT_RANDOM_NUMBER_SEED,
DEFAULT_TRUE_NUMBER_OF_CLASSES)
distinctTrainLabels = np.unique(labels)
OPACITY = 0.7
plt.figure()
plt.title("Data Set")
plt.xlabel("Feature 1")
plt.ylabel("Feature 2")
for i, label in enumerate(distinctTrainLabels):
plt.scatter(data[labels == label, 0],
data[labels == label, 1],
c=np.random.rand(3, ),
alpha=OPACITY,
label="Class {}".format(i))
plt.legend()
plt.show()
from scipy.optimize import curve_fit
import numpy as np
from pylab import *
from DataGenerator import generateData
def mySin(x, a=1, b=1):
return a*sin(b*x)
def fit(xValues, yValues):
popt, pcov = curve_fit(mySin, xValues, yValues)
yValuesFit = [ mySin(xV, popt[0], popt[1]) for xV in xValues ]
return yValuesFit
if __name__=='__main__':
xValues = arange(0, 7, 0.1)
yValues = generateData( xValues, [-2, 2])
yValuesFit = fit(xValues, yValues)
figure(1)
plot(xValues, yValues, 'go')
plot(xValues, yValuesFit, 'b')
sinValues = [ sin(x) for x in xValues]
plot(xValues, sinValues, 'r')
grid(True)
show()