def eval_prox(prox_function_type, prob, v_map, lam=1, epigraph=False):
"""Evaluate a single proximal operator."""
problem = compiler.compile_problem(cvxpy_expr.convert_problem(prob))
validate.check_sum_of_prox(problem)
# Get the first non constant objective term
if len(problem.objective.arg) != 1:
raise ProblemError("prox does not have single f", problem)
if problem.constraint:
raise ProblemError("prox has constraints", problem)
f_expr = problem.objective.arg[0]
if (f_expr.expression_type != Expression.PROX_FUNCTION or
f_expr.prox_function.prox_function_type != prox_function_type or
f_expr.prox_function.epigraph != epigraph):
raise ProblemError("prox did not compile to right type", problem)
v_bytes_map = {cvxpy_expr.variable_id(var):
numpy.array(val, dtype=numpy.float64).tobytes(order="F")
for var, val in v_map.iteritems()}
values = _solve.eval_prox(
f_expr.SerializeToString(),
lam,
constant.global_data_map,
v_bytes_map)
cvxpy_solver.set_solution(prob, values)
def eval_prox_impl(prob,
v_map,
lam=1,
prox_function_type=None,
epigraph=False):
"""Evaluate a single proximal operator."""
problem = compiler.compile_problem(cvxpy_expr.convert_problem(prob))
validate.check_sum_of_prox(problem)
# Get the first non constant objective term
if len(problem.objective.arg) != 1:
raise ProblemError("prox does not have single f", problem)
if problem.constraint:
raise ProblemError("prox has constraints", problem)
f_expr = problem.objective.arg[0]
if f_expr.expression_type != Expression.PROX_FUNCTION:
raise ProblemError("prox did not compile to right type", problem)
if prox_function_type is not None and (
f_expr.prox_function.prox_function_type != prox_function_type
or f_expr.prox_function.epigraph != epigraph):
raise ProblemError("prox did not compile to right type", problem)
v_bytes_map = {
cvxpy_expr.variable_id(var):
numpy.array(val, dtype=numpy.float64).tobytes(order="F")
for var, val in v_map.iteritems()
}
values = _solve.eval_prox(f_expr.SerializeToString(), lam,
problem.expression_data(), v_bytes_map)
cvxpy_solver.set_solution(prob, values)
def set_solution(prob, values):
for var in prob.variables():
var_id = cvxpy_expr.variable_id(var)
assert var_id in values
x = numpy.fromstring(values[var_id], dtype=numpy.double)
var.value = x.reshape(var.size[1], var.size[0]).transpose()
def set_solution(prob, values):
for var in prob.variables():
var_id = cvxpy_expr.variable_id(var)
assert var_id in values
x = numpy.fromstring(values[var_id], dtype=numpy.double)
var.value = x.reshape(var.size[1], var.size[0]).transpose()
