def test_objective_to_coeffs():
vars = [x, y, z]
assert_array_equal(objective_to_coeffs(1.2 + x - 3.4 * y, vars, 'max'),
array([-1.0, 3.4, 0.0]))
except_ok = False
try:
objective_to_coeffs(1.2 + x * y, vars)
except ValueError:
except_ok = True
assert except_ok
def to_sdpa_sparse(objective_func,
lmis,
variables,
objective_type='minimize',
split_blocks=True,
comment=None):
"""Put problem (objective and LMIs) into SDPA sparse format."""
obj_coeffs = lmi_sdp.objective_to_coeffs(objective_func, variables,
objective_type)
lmi_coeffs = lmi_to_coeffs(lmis, variables, split_blocks)
s = lmi_sdp.sdp._sdpa_header(obj_coeffs, lmi_coeffs, comment)
def _print_sparse(x, b, m, sign=1):
s = ''
shape = m.shape
for i in range(shape[0]):
for j in range(i, shape[1]):
e = sign * m[i, j]
if e != 0:
s += '%d %d %d %d %s\n' % (x, b, i + 1, j + 1, str(e))
return s
for b in range(len(lmi_coeffs)):
s += _print_sparse(0, b + 1, lmi_coeffs[b][1], sign=-1)
for x in range(len(obj_coeffs)):
for b in range(len(lmi_coeffs)):
s += _print_sparse(x + 1, b + 1, lmi_coeffs[b][0][x])
return s
def to_sdpa_sparse(objective_func, lmis, variables, objective_type='minimize',
split_blocks=True, comment=None):
"""Put problem (objective and LMIs) into SDPA sparse format."""
obj_coeffs = lmi_sdp.objective_to_coeffs(objective_func, variables,
objective_type)
lmi_coeffs = lmi_to_coeffs(lmis, variables, split_blocks)
s = lmi_sdp.sdp._sdpa_header(obj_coeffs, lmi_coeffs, comment)
def _print_sparse(x, b, m, sign=1):
s = ''
shape = m.shape
for i in range(shape[0]):
for j in range(i, shape[1]):
e = sign*m[i, j]
if e != 0:
s += '%d %d %d %d %s\n' % (x, b, i+1, j+1, str(e))
return s
for b in range(len(lmi_coeffs)):
s += _print_sparse(0, b+1, lmi_coeffs[b][1], sign=-1)
for x in range(len(obj_coeffs)):
for b in range(len(lmi_coeffs)):
s += _print_sparse(x+1, b+1, lmi_coeffs[b][0][x])
return s
def to_cvxopt(objective_func,
lmis,
variables,
objective_type='minimize',
split_blocks=True):
"""Prepare objective and LMI to be used with cvxopt SDP solver.
Parameters
----------
objective_func: symbolic linear expression
lmi: symbolic LMI or Matrix, or a list of them
variables: list of symbols
The variable symbols which form the LMI/SDP space.
objective_type: 'maximize' or 'minimize', defaults to 'minimize'
split_blocks: bool
If set to True, function tries to subdivide each LMI into
smaller diagonal blocks
Returns
-------
c, Gs, hs: parameters ready to be input to cvxopt.solvers.sdp()
"""
if cvxopt is None:
raise lmi_sdp.sdp.NotAvailableError(to_cvxopt.__name__)
obj_coeffs = lmi_sdp.objective_to_coeffs(objective_func, variables,
objective_type)
lmi_coeffs = lmi_to_coeffs(lmis, variables, split_blocks)
c = matrix(obj_coeffs)
Gs = []
hs = []
for (LMis, LM0) in lmi_coeffs:
Gs.append(
matrix([(-LMi).flatten().astype(float).tolist() for LMi in LMis]))
hs.append(matrix(LM0.astype(float).tolist()))
return c, Gs, hs
def to_cvxopt(objective_func, lmis, variables, objective_type='minimize',
split_blocks=True):
"""Prepare objective and LMI to be used with cvxopt SDP solver.
Parameters
----------
objective_func: symbolic linear expression
lmi: symbolic LMI or Matrix, or a list of them
variables: list of symbols
The variable symbols which form the LMI/SDP space.
objective_type: 'maximize' or 'minimize', defaults to 'minimize'
split_blocks: bool
If set to True, function tries to subdivide each LMI into
smaller diagonal blocks
Returns
-------
c, Gs, hs: parameters ready to be input to cvxopt.solvers.sdp()
"""
if cvxopt is None:
raise NotAvailableError(to_cvxopt.__name__)
obj_coeffs = lmi_sdp.objective_to_coeffs(objective_func, variables,
objective_type)
lmi_coeffs = lmi_to_coeffs(lmis, variables, split_blocks)
c = matrix(obj_coeffs)
Gs = []
hs = []
for (LMis, LM0) in lmi_coeffs:
Gs.append(matrix([(-LMi).flatten().astype(float).tolist()
for LMi in LMis]))
hs.append(matrix(LM0.astype(float).tolist()))
return c, Gs, hs
