def upper_tri(n):
data = {}
m = n * (n - 1) / 2
rows = np.arange(m)
cols = np.array([j * n + i for i in xrange(n) for j in xrange(i + 1, n)])
A = sp.coo_matrix((np.ones(m), (rows, cols)), shape=(m, n * n))
return sparse_matrix(constant.store(A, data), data)
def transpose_matrix(m, n):
A = sp.coo_matrix(
(np.ones(m*n),
(np.arange(m*n),
np.tile(np.arange(n)*m, m) + np.repeat(np.arange(m), n))),
shape=(m*n, m*n))
return sparse_matrix(constant.store(A))
def ones(*dims):
data = {}
return Expression(expression_type=expression_pb2.Expression.CONSTANT,
size=Size(dim=dims),
data=data,
constant=_constant.store(np.ones(dims), data),
func_curvature=CONSTANT)
def ones(*dims):
data = {}
return Expression(
expression_type=expression_pb2.Expression.CONSTANT,
size=Size(dim=dims),
data=data,
constant=_constant.store(np.ones(dims), data),
func_curvature=CONSTANT)
def index(slice, n):
m = slice.stop - slice.start
if m == n:
return identity(n)
A = sp.coo_matrix(
(np.ones(m),
(np.arange(m), np.arange(slice.start, slice.stop, slice.step))),
shape=(m, n))
return sparse_matrix(constant.store(A))
def convert_constant(expr):
m, n = expr.size
if isinstance(expr.value, (int, long, float)):
return expression.constant(m, n, scalar=expr.value)
else:
return expression.constant(
m, n,
constant=constant.store(expr.value),
sign=expression_pb2.Sign(
sign_type=expression_pb2.Sign.Type.Value(expr.sign)))
def index(slice, n):
m = slice.stop - slice.start
if m == n:
return identity(n)
A = sp.coo_matrix(
(np.ones(m),
(np.arange(m), np.arange(slice.start, slice.stop, slice.step))),
shape=(m, n))
data = {}
return sparse_matrix(constant.store(A, data), data)
def convert_constant(expr):
m, n = expr.size
if isinstance(expr.value, (int, float)):
return expression.constant(m, n, scalar=expr.value)
else:
data = {}
return expression.constant(
m, n,
data=data,
constant=constant.store(expr.value, data),
sign=expression_pb2.Sign(
sign_type=expression_pb2.Sign.Type.Value(expr.sign)))
def sum_left(m, n):
data = {}
return left_matrix_product(
dense_matrix(constant.store(np.ones((1, m)), data), data), n)
def sum_right(m, n):
return right_matrix_product(dense_matrix(constant.store(np.ones((n,1)))), m)
def sum_left(m, n):
return left_matrix_product(dense_matrix(constant.store(np.ones((1,m)))), n)
def sum(m, n):
return kronecker_product(
dense_matrix(constant.store(np.ones((1,n)))),
dense_matrix(constant.store(np.ones((1,m)))))
def transpose_matrix(m, n):
A = sp.coo_matrix((np.ones(m * n), (np.arange(
m * n), np.tile(np.arange(n) * m, m) + np.repeat(np.arange(m), n))),
shape=(m * n, m * n))
data = {}
return sparse_matrix(constant.store(A, data), data)
def upper_tri(n):
m = n*(n-1)/2
rows = np.arange(m)
cols = np.array([j*n + i for i in xrange(n) for j in xrange(i+1,n)])
A = sp.coo_matrix((np.ones(m), (rows, cols)), shape=(m, n*n))
return sparse_matrix(constant.store(A))
def one_hot(i, n):
return sparse_matrix(
constant.store(sp.coo_matrix(([1], ([i], [0])), shape=(n,1))))
def sum_right(m, n):
data = {}
return right_matrix_product(
dense_matrix(constant.store(np.ones((n, 1)), data), data), m)
def trace(n):
data = {}
rows = np.zeros(n)
cols = np.arange(n) * (n + 1)
A = sp.coo_matrix((np.ones(n), (rows, cols)), shape=(1, n * n))
return sparse_matrix(constant.store(A, data), data)
def diag_vec(n):
data = {}
rows = np.arange(n) * (n + 1)
cols = np.arange(n)
A = sp.coo_matrix((np.ones(n), (rows, cols)), shape=(n * n, n))
return sparse_matrix(constant.store(A, data), data)
def diag_vec(n):
rows = np.arange(n)*(n+1)
cols = np.arange(n)
A = sp.coo_matrix((np.ones(n), (rows, cols)), shape=(n*n, n))
return sparse_matrix(constant.store(A))
def trace(n):
rows = np.zeros(n)
cols = np.arange(n)*(n+1)
A = sp.coo_matrix((np.ones(n), (rows, cols)), shape=(1, n*n))
return sparse_matrix(constant.store(A))
def promote(n):
return dense_matrix(constant.store(np.ones((n,1))))
def sum(m, n):
data = {}
return kronecker_product(
dense_matrix(constant.store(np.ones((1, n)), data), data),
dense_matrix(constant.store(np.ones((1, m)), data), data))
def promote(n):
data = {}
return dense_matrix(constant.store(np.ones((n, 1)), data), data)
def sum(n):
return dense_matrix(constant.store(np.ones((1,n))))
def parameter_values(cvxpy_prob, data):
return [(cvxpy_expr.parameter_id(param),
constant.store(param.value, data).SerializeToString())
for param in cvxpy_prob.parameters()]
def parameter_values(cvxpy_prob, data):
return [(cvxpy_expr.parameter_id(param),
constant.store(param.value, data).SerializeToString())
for param in cvxpy_prob.parameters()]
def one_hot(i, n):
return sparse_matrix(
constant.store(sp.coo_matrix(([1], ([i], [0])), shape=(n, 1))))
