def compute_image(self):
image = ModVectorSpace([])
for row in self.get_A().get_rref():
if not row.is_zero():
index = row.get_leading_index()
image.add_basis_element([ self.get_A().get_column(index) ])
self.image = image
self.image_flag = True
def __init__(self, B, A):
self.B = B
self.A = A
self.kernel = None
self.kernel_flag = False
self.image = None
self.image_flag = False
self.cohomology = None
self.cohomology_flag = False
self.product = ModVectorSpace([])
# if self.A.rows == []:
# self.A = ModMatrix.null(self.B.get_size()[1], 1)
# elif self.B.rows == []:
# raise TypeError
if self.B.get_size()[1] != self.A.get_size()[0]:
raise TypeError("wrong sizes B: " + str(self.B) + "A: " +
str(self.A))
self.extended_ker_basis = None
self.extended_ker_basis_flag = False
def compute_cohomology(self):
cohomology = ModVectorSpace([])
##### Tweak to check dimension before working
if not self.get_cohomology_dim():
self.cohomology = cohomology
self.cohomology_flag = True
return
#####
index_list = self.get_kernel().get_basis()[:]
basis_list = []
for ker_element in index_list:
if not self.in_cohomology_subspace(basis_list, ker_element):
basis_list.append(ker_element)
cohomology.add_basis_element(basis_list)
self.cohomology = cohomology
if (self.get_B().col_count - self.get_B().get_rank()
- self.get_A().get_rank() != len(cohomology.get_basis())):
print "A"
print self.get_A()
print "B"
print self.get_B()
print "rank A"
print self.get_A().get_rank()
print "nullity B"
print self.get_B().col_count - self.get_B().get_rank()
print "basis for cohomology"
for thing in cohomology.get_basis():
print thing
raise TypeError("cohomology basis computation incorrect!")
self.cohomology_flag = True
def get_kernel(self):
if not self.get_kernel_flag():
self.kernel = ModVectorSpace(self.get_B().get_kernel_vout())
self.kernel_flag = True
return self.kernel
class Cohomology(object):
"""This class defines a cohomology object. It is initialized with two
ModMatrix objects B, A such that B*A is defined. We assume that
both matrices are expressed in terms of a common basis in
codomain(A) = domain(B).
This class has methods which will determine a basis for Ker(B), a
basis for Im(A), and for Ker(B)/Im(A). There is a method to
determine if two cocycles are equivalent in cohomology.
Now will have an attribute to store a basis for cohomology in
terms of product structure.
"""
def __init__(self, B, A):
self.B = B
self.A = A
self.kernel = None
self.kernel_flag = False
self.image = None
self.image_flag = False
self.cohomology = None
self.cohomology_flag = False
self.product = ModVectorSpace([])
# if self.A.rows == []:
# self.A = ModMatrix.null(self.B.get_size()[1], 1)
# elif self.B.rows == []:
# raise TypeError
if self.B.get_size()[1] != self.A.get_size()[0]:
raise TypeError("wrong sizes B: " + str(self.B) + "A: " +
str(self.A))
self.extended_ker_basis = None
self.extended_ker_basis_flag = False
def get_A(self):
return self.A
def get_B(self):
return self.B
def get_kernel_flag(self):
return self.kernel_flag
def get_image_flag(self):
return self.image_flag
def get_cohomology_flag(self):
return self.cohomology_flag
def get_cohomology_dim(self):
return self.B.get_nullity() - self.A.get_rank()
def get_zero_vector(self):
dim = self.get_A().get_size()[0]
return ModVector.null(dim)
def im_in_ker(self):
C = self.get_B() * self.get_A()
if not C.is_zero():
return False
else:
return True
def get_kernel(self):
if not self.get_kernel_flag():
self.kernel = ModVectorSpace(self.get_B().get_kernel_vout())
self.kernel_flag = True
return self.kernel
def in_kernel(self, vect):
"""
Is vect a ModMatrix or a ModVector??? Fix this
"""
C = self.get_B() * vect
return C.is_zero()
def compute_image(self):
image = ModVectorSpace([])
for row in self.get_A().get_rref():
if not row.is_zero():
index = row.get_leading_index()
image.add_basis_element([ self.get_A().get_column(index) ])
self.image = image
self.image_flag = True
def get_image(self):
if not self.get_image_flag():
self.compute_image()
return self.image
def are_cohomologous(self, vect_1, vect_2):
"""
inputs are bitvectors, not bitmatrix object representing col. vector
"""
vector_sum = vect_1 + ( (-1) * vect_2)
return self.get_A().can_solve(vector_sum)
def compute_cohomology(self):
cohomology = ModVectorSpace([])
##### Tweak to check dimension before working
if not self.get_cohomology_dim():
self.cohomology = cohomology
self.cohomology_flag = True
return
#####
index_list = self.get_kernel().get_basis()[:]
basis_list = []
for ker_element in index_list:
if not self.in_cohomology_subspace(basis_list, ker_element):
basis_list.append(ker_element)
cohomology.add_basis_element(basis_list)
self.cohomology = cohomology
if (self.get_B().col_count - self.get_B().get_rank()
- self.get_A().get_rank() != len(cohomology.get_basis())):
print "A"
print self.get_A()
print "B"
print self.get_B()
print "rank A"
print self.get_A().get_rank()
print "nullity B"
print self.get_B().col_count - self.get_B().get_rank()
print "basis for cohomology"
for thing in cohomology.get_basis():
print thing
raise TypeError("cohomology basis computation incorrect!")
self.cohomology_flag = True
def get_cohomology(self):
if not self.get_cohomology_flag():
#Bravely skip this to save time. hope for the best?
# if not self.im_in_ker():
# print "matrix for image"
# print str(self.A)
# print "matrix for kernel"
# print str(self.B)
# raise TypeError("Not exact here!")
self.compute_cohomology()
return self.cohomology
def get_product(self):
return self.product
def add_to_product_basis(self, a_list):
"""
input is a list of bitvectors
"""
self.product.add_basis_element(a_list)
def in_cohomology_subspace(self, subspace_basis, vector):
"""
subspace_basis should be given as bitvectors
vector is also a bit vector
all should be in terms of standard basis for
codomain(A) = domain(B)
"""
B = self.get_A().get_append_columns(subspace_basis)
# print "checking if\n", vector, "\nin subspace\n", B
# print "result", B.can_solve(vector)
return B.can_solve(vector)
# def coset(self, vector):
# """
# given a vector V, returns the set
# V + Im(A)
# """
# out_set = set()
# im = self.get_image().get_basis()
# coeffs = BitVector.vector_space_set(len(im))
# for cc in coeffs:
# new = vector.copy()
# for i in range(0,len(im)):
# if cc[i]:
# new += im[i]
# out_set.add(new)
# return out_set
# def short_representative(self, vector):
# coset = self.coset(vector)
# min_cpt = vector.number_of_components()
# out = vector
# for w in coset:
# if w.number_of_components() < min_cpt:
# min_cpt = w.number_of_components()
# out = w
# return out
def extend_ker_basis(self):
ker_basis = self.get_kernel().get_basis()
#assuming vectors are bit vectors
if not ker_basis:
C = ModMatrix.null(1, self.B.col_count)
else:
C = ModMatrix(ker_basis)
comp = C.complement_row_space()
tot = ker_basis[:] + comp[:]
self.extended_ker_basis = tot
self.extended_ker_basis_flag = True
def get_extended_ker_basis(self):
if not self.extended_ker_basis_flag:
self.extend_ker_basis()
return self.extended_ker_basis
def basis_to_ker_basis(self):
C = ModMatrix(self.get_extended_ker_basis())
C.transpose()
inv = C.get_inverse()
if not inv:
raise TypeError("the matrix is not invertible! Something went very wrong")
return inv
