def fit(self, epoch=10):
A = self.Mat
data = A.data
X = self.X
while epoch:
m = absmax(X)
X = [i / m for i in X]
X = GaussEq(A, X)
epoch -= 1
AX = [sum([A_i[j] * X_j for j, X_j in enumerate(X)]) for A_i in data]
m1 = sum(dot_vec(AX, X)) / sum(dot_vec(X, X))
self.X = X
return m1
def fit(self, B):
if isinstance(B, matrix):
B = B.c(range(B.size[0]), 0)
n = len(B)
L = self.L
U = self.U
y = [0 for i in xrange(n)]
x = [0 for i in xrange(n)]
for i in xrange(n):
rang = range(i)
y[i] = B[i] - sum(dot_vec(L.c([i], rang), y[:i]))
for i in xrange(n - 1, -1, -1):
rang = range(i + 1, n)
x[i] = (y[i] -
sum(dot_vec(U.c([i], rang), x[(i + 1):]))) / U.data[i][i]
return x
def GaussEq(A, B):
if isinstance(B, matrix):
B = B.c(range(B.size[0]), 0)
B = copy.deepcopy(B)
size = A.size
data = copy.deepcopy(A.data)
row_arg = range(size[0])
if size[0] != size[1]:
print('error\n')
return ValueError
size = size[0]
for i in row_arg[:-1]:
vec = matrix(data).c(range(i, size), i)
max_index = i + argabsmax(vec)
data[max_index], data[i] = data[i], data[max_index]
B[max_index], B[i] = B[i], B[max_index]
for index in range(i + 1, size):
try:
param = data[index][i] / data[i][i]
data[index] = add_vec(data[index], data[i], -param)
B[index] += B[i] * (-param)
except:
print('error: a singular matrix')
return ValueError
row_arg.reverse()
x = [0 for i in range(size)]
for i in row_arg:
x[i] = (B[i] - sum(dot_vec(x[i:], data[i][i:]))) / data[i][i]
return x
def fit(self, epoch=10):
data = self.Mat.data
X = self.X
while epoch:
m = absmax(X)
X = [i / m for i in X]
X = [
sum([A_i[j] * X_j for j, X_j in enumerate(X)]) for A_i in data
]
epoch -= 1
#get rayleigh div
AX = [sum([A_i[j] * X_j for j, X_j in enumerate(X)]) for A_i in data]
m1 = sum(dot_vec(AX, X)) / sum(dot_vec(X, X))
#m2=absmax(X)
self.X = X
return m1 #,m2
def LU(A):
size = A.size
if size[0] != size[1]:
return ValueError
L = matrix(size, eye=True)
U = matrix(size, eye=False)
size = size[0]
aij = A.data
for i in range(0, size):
rang = range(0, i)
for r in range(i, size):
U.data[i][r] = aij[i][r] - sum(
dot_vec(L.c([i], rang), U.c(rang, [r])))
for r in range(i + 1, size):
L.data[r][i] = (aij[r][i] - sum(
dot_vec(L.c([r], rang), U.c(rang, [i])))) / U.data[i][i]
return L.data, U.data
def Chol(A):
size = A.size
if size[0] != size[1]:
return ValueError
if not check_symmetrical(A):
print 'Not symmetrical!\n'
return ValueError
aij = A.data
L = matrix(size, eye=False)
size = size[0]
for j in range(size):
try:
L[j][j] = math.sqrt(aij[j][j] - sum(dot_vec(L[j][:j], L[j][:j])))
except:
print 'error:Not positive!\n'
for i in range(j + 1, size):
try:
L[i][j] = (aij[i][j] -
sum(dot_vec(L[i][:j], L[j][:j]))) / L[j][j]
except:
print 'error:Not positive!\n'
return L
def vecmean(vecs):
summa = reduce(add_vec, vecs)
n = len(vecs)
ret = dot_vec(1 / n, summa)
return ret