Source code for pyomo.core.kernel.piecewise_library.transforms

#  ___________________________________________________________________________
#
#  Pyomo: Python Optimization Modeling Objects
#  Copyright (c) 2008-2024
#  National Technology and Engineering Solutions of Sandia, LLC
#  Under the terms of Contract DE-NA0003525 with National Technology and
#  Engineering Solutions of Sandia, LLC, the U.S. Government retains certain
#  rights in this software.
#  This software is distributed under the 3-clause BSD License.
#  ___________________________________________________________________________

"""
This module contains transformations for representing a
single-variate piecewise linear function using a
mixed-integer problem formulation (see [VAN10]_).

"""

import logging
import bisect

# TODO: Figure out of the 'log' and 'dlog' representations
# really do require (2^n)+1 points or if there is a way to
# handle the between sizes.

from pyomo.core.expr.numvalue import value as _value
from pyomo.core.kernel.set_types import IntegerSet
from pyomo.core.kernel.block import block
from pyomo.core.kernel.expression import expression, expression_tuple
from pyomo.core.kernel.variable import (
    IVariable,
    variable_list,
    variable_tuple,
    variable_dict,
    variable,
)
from pyomo.core.kernel.constraint import (
    constraint_list,
    constraint_tuple,
    linear_constraint,
)
from pyomo.core.kernel.sos import sos2
from pyomo.core.kernel.piecewise_library.util import (
    characterize_function,
    is_nondecreasing,
    is_positive_power_of_two,
    log2floor,
    generate_gray_code,
    PiecewiseValidationError,
)


logger = logging.getLogger('pyomo.core')

registered_transforms = {}


# wrapper that allows a list containing parameters to be
# used with the bisect module
class _shadow_list(object):
    __slots__ = ("_x",)

    def __init__(self, x):
        self._x = x

    def __len__(self):
        return self._x.__len__()

    def __getitem__(self, i):
        return _value(self._x.__getitem__(i))


[docs] def piecewise( breakpoints, values, input=None, output=None, bound='eq', repn='sos2', validate=True, simplify=True, equal_slopes_tolerance=1e-6, require_bounded_input_variable=True, require_variable_domain_coverage=True, ): """ Models a single-variate piecewise linear function. This function takes a list breakpoints and function values describing a piecewise linear function and transforms this input data into a block of variables and constraints that enforce a piecewise linear relationship between an input variable and an output variable. In the general case, this transformation requires the use of discrete decision variables. Args: breakpoints (list): The list of breakpoints of the piecewise linear function. This can be a list of numbers or a list of objects that store mutable data (e.g., mutable parameters). If mutable data is used validation might need to be disabled by setting the :attr:`validate` keyword to :const:`False`. The list of breakpoints must be in non-decreasing order. values (list): The values of the piecewise linear function corresponding to the breakpoints. input: The variable constrained to be the input of the piecewise linear function. output: The variable constrained to be the output of the piecewise linear function. bound (str): The type of bound to impose on the output expression. Can be one of: - 'lb': y <= f(x) - 'eq': y = f(x) - 'ub': y >= f(x) repn (str): The type of piecewise representation to use. Choices are shown below (+ means step functions are supported) - 'sos2': standard representation using sos2 constraints (+) - 'dcc': disaggregated convex combination (+) - 'dlog': logarithmic disaggregated convex combination (+) - 'cc': convex combination (+) - 'log': logarithmic branching convex combination (+) - 'mc': multiple choice - 'inc': incremental method (+) validate (bool): Indicates whether or not to perform validation of the input data. The default is :const:`True`. Validation can be performed manually after the piecewise object is created by calling the :meth:`validate` method. Validation should be performed any time the inputs are changed (e.g., when using mutable parameters in the breakpoints list or when the input variable changes). simplify (bool): Indicates whether or not to attempt to simplify the piecewise representation to avoid using discrete variables. This can be done when the feasible region for the output variable, with respect to the piecewise function and the bound type, is a convex set. Default is :const:`True`. Validation is required to perform simplification, so this keyword is ignored when the :attr:`validate` keyword is :attr:`False`. equal_slopes_tolerance (float): Tolerance used check if consecutive slopes are nearly equal. If any are found, validation will fail. Default is 1e-6. This keyword is ignored when the :attr:`validate` keyword is :attr:`False`. require_bounded_input_variable (bool): Indicates if the input variable is required to have finite upper and lower bounds. Default is :const:`True`. Setting this keyword to :const:`False` can be used to allow general expressions to be used as the input in place of a variable. This keyword is ignored when the :attr:`validate` keyword is :attr:`False`. require_variable_domain_coverage (bool): Indicates if the function domain (defined by the endpoints of the breakpoints list) needs to cover the entire domain of the input variable. Default is :const:`True`. Ignored for any bounds of variables that are not finite, or when the input is not assigned a variable. This keyword is ignored when the :attr:`validate` keyword is :attr:`False`. Returns: TransformedPiecewiseLinearFunction: a block that stores any new variables, constraints, and other modeling objects used by the piecewise representation """ transform = None try: transform = registered_transforms[repn] except KeyError: raise ValueError( "Keyword assignment repn='%s' is not valid. " "Must be one of: %s" % (repn, str(sorted(registered_transforms.keys()))) ) assert transform is not None if not validate: # can not simplify if we do not validate simplify = False func = PiecewiseLinearFunction(breakpoints, values, validate=False) if simplify and (transform is not piecewise_convex): ftype = func.validate(equal_slopes_tolerance=equal_slopes_tolerance) if (bound == 'eq') and (ftype == characterize_function.affine): transform = piecewise_convex elif (bound == 'lb') and ( ftype in (characterize_function.affine, characterize_function.convex) ): transform = piecewise_convex elif (bound == 'ub') and ( ftype in (characterize_function.affine, characterize_function.concave) ): transform = piecewise_convex return transform( func, input=input, output=output, bound=bound, validate=validate, equal_slopes_tolerance=equal_slopes_tolerance, require_bounded_input_variable=require_bounded_input_variable, require_variable_domain_coverage=require_variable_domain_coverage, )
[docs] class PiecewiseLinearFunction(object): """A piecewise linear function Piecewise linear functions are defined by a list of breakpoints and a list function values corresponding to each breakpoint. The function value between breakpoints is implied through linear interpolation. Args: breakpoints (list): The list of function breakpoints. values (list): The list of function values (one for each breakpoint). validate (bool): Indicates whether or not to perform validation of the input data. The default is :const:`True`. Validation can be performed manually after the piecewise object is created by calling the :meth:`validate` method. Validation should be performed any time the inputs are changed (e.g., when using mutable parameters in the breakpoints list). **kwds: Additional keywords are passed to the :meth:`validate` method when the :attr:`validate` keyword is :const:`True`; otherwise, they are ignored. """ __slots__ = ("_breakpoints", "_values")
[docs] def __init__(self, breakpoints, values, validate=True, **kwds): self._breakpoints = breakpoints self._values = values if type(self._breakpoints) is not tuple: self._breakpoints = tuple(self._breakpoints) if type(self._values) is not tuple: self._values = tuple(self._values) if len(self._breakpoints) != len(self._values): raise ValueError( "The number of breakpoints (%s) differs from " "the number of function values (%s)" % (len(self._breakpoints), len(self._values)) ) if validate: self.validate(**kwds)
def __getstate__(self): """Required for older versions of the pickle protocol since this class uses __slots__""" return {key: getattr(self, key) for key in self.__slots__} def __setstate__(self, state): """Required for older versions of the pickle protocol since this class uses __slots__""" for key in state: setattr(self, key, state[key])
[docs] def validate(self, equal_slopes_tolerance=1e-6): """ Validate this piecewise linear function by verifying various properties of the breakpoints and values lists (e.g., that the list of breakpoints is nondecreasing). Args: equal_slopes_tolerance (float): Tolerance used check if consecutive slopes are nearly equal. If any are found, validation will fail. Default is 1e-6. Returns: int: a function characterization code (see :func:`util.characterize_function`) Raises: PiecewiseValidationError: if validation fails """ breakpoints = [_value(x) for x in self._breakpoints] values = [_value(x) for x in self._values] if not is_nondecreasing(breakpoints): raise PiecewiseValidationError( "The list of breakpoints is not nondecreasing: %s" % (str(breakpoints)) ) ftype, slopes = characterize_function(breakpoints, values) for i in range(1, len(slopes)): if ( (slopes[i - 1] is not None) and (slopes[i] is not None) and (abs(slopes[i - 1] - slopes[i]) <= equal_slopes_tolerance) ): raise PiecewiseValidationError( "Piecewise function validation detected slopes " "of consecutive line segments to be within %s " "of one another. This may cause numerical issues. " "To avoid this error, set the 'equal_slopes_tolerance' " "keyword to a smaller value or disable validation." % (equal_slopes_tolerance) ) return ftype
@property def breakpoints(self): """The set of breakpoints used to defined this function""" return self._breakpoints @property def values(self): """The set of values used to defined this function""" return self._values def __call__(self, x): """Evaluates the piecewise linear function at the given point using interpolation. Note that step functions are assumed lower-semicontinuous.""" i = bisect.bisect_left(_shadow_list(self.breakpoints), x) if i == 0: xP = _value(self.breakpoints[i]) if xP == x: return float(_value(self.values[i])) elif i != len(self.breakpoints): xL = _value(self.breakpoints[i - 1]) xU = _value(self.breakpoints[i]) assert xL <= xU if (xL <= x) and (x <= xU): yL = _value(self.values[i - 1]) yU = _value(self.values[i]) return yL + (float(yU - yL) / (xU - xL)) * (x - xL) raise ValueError( "The point %s is outside of the " "function domain: [%s,%s]." % (x, _value(self.breakpoints[0]), _value(self.breakpoints[-1])) )
[docs] class TransformedPiecewiseLinearFunction(block): """Base class for transformed piecewise linear functions A transformed piecewise linear functions is a block of variables and constraints that enforce a piecewise linear relationship between an input variable and an output variable. Args: f (:class:`PiecewiseLinearFunction`): The piecewise linear function to transform. input: The variable constrained to be the input of the piecewise linear function. output: The variable constrained to be the output of the piecewise linear function. bound (str): The type of bound to impose on the output expression. Can be one of: - 'lb': y <= f(x) - 'eq': y = f(x) - 'ub': y >= f(x) validate (bool): Indicates whether or not to perform validation of the input data. The default is :const:`True`. Validation can be performed manually after the piecewise object is created by calling the :meth:`validate` method. Validation should be performed any time the inputs are changed (e.g., when using mutable parameters in the breakpoints list or when the input variable changes). **kwds: Additional keywords are passed to the :meth:`validate` method when the :attr:`validate` keyword is :const:`True`; otherwise, they are ignored. """
[docs] def __init__(self, f, input=None, output=None, bound='eq', validate=True, **kwds): super(TransformedPiecewiseLinearFunction, self).__init__() assert isinstance(f, PiecewiseLinearFunction) if bound not in ('lb', 'ub', 'eq'): raise ValueError( "Invalid bound type %r. Must be one of: ['lb','ub','eq']" % (bound) ) self._bound = bound self._f = f self._inout = expression_tuple([expression(input), expression(output)]) if validate: self.validate(**kwds)
@property def input(self): """The expression that stores the input to the piecewise function. The returned object can be updated by assigning to its :attr:`expr` attribute.""" return self._inout[0] @property def output(self): """The expression that stores the output of the piecewise function. The returned object can be updated by assigning to its :attr:`expr` attribute.""" return self._inout[1] @property def bound(self): """The bound type assigned to the piecewise relationship ('lb','ub','eq').""" return self._bound
[docs] def validate( self, equal_slopes_tolerance=1e-6, require_bounded_input_variable=True, require_variable_domain_coverage=True, ): """ Validate this piecewise linear function by verifying various properties of the breakpoints, values, and input variable (e.g., that the list of breakpoints is nondecreasing). Args: equal_slopes_tolerance (float): Tolerance used check if consecutive slopes are nearly equal. If any are found, validation will fail. Default is 1e-6. require_bounded_input_variable (bool): Indicates if the input variable is required to have finite upper and lower bounds. Default is :const:`True`. Setting this keyword to :const:`False` can be used to allow general expressions to be used as the input in place of a variable. require_variable_domain_coverage (bool): Indicates if the function domain (defined by the endpoints of the breakpoints list) needs to cover the entire domain of the input variable. Default is :const:`True`. Ignored for any bounds of variables that are not finite, or when the input is not assigned a variable. Returns: int: a function characterization code (see :func:`util.characterize_function`) Raises: PiecewiseValidationError: if validation fails """ ftype = self._f.validate(equal_slopes_tolerance=equal_slopes_tolerance) assert ftype in (1, 2, 3, 4, 5) input_var = self.input.expr if not isinstance(input_var, IVariable): input_var = None if require_bounded_input_variable and ( (input_var is None) or (not input_var.has_lb()) or (not input_var.has_ub()) ): raise PiecewiseValidationError( "Piecewise function input is not a " "variable with finite upper and lower " "bounds: %s. To avoid this error, set the " "'require_bounded_input_variable' keyword " "to False or disable validation." % (str(input_var)) ) if require_variable_domain_coverage and (input_var is not None): domain_lb = _value(self.breakpoints[0]) domain_ub = _value(self.breakpoints[-1]) if input_var.has_lb() and _value(input_var.lb) < domain_lb: raise PiecewiseValidationError( "Piecewise function domain does not include " "the lower bound of the input variable: " "%s.ub = %s > %s. To avoid this error, set " "the 'require_variable_domain_coverage' " "keyword to False or disable validation." % (input_var.name, _value(input_var.lb), domain_lb) ) if input_var.has_ub() and _value(input_var.ub) > domain_ub: raise PiecewiseValidationError( "Piecewise function domain does not include " "the upper bound of the input variable: " "%s.ub = %s > %s. To avoid this error, set " "the 'require_variable_domain_coverage' " "keyword to False or disable validation." % (input_var.name, _value(input_var.ub), domain_ub) ) return ftype
@property def breakpoints(self): """The set of breakpoints used to defined this function""" return self._f.breakpoints @property def values(self): """The set of values used to defined this function""" return self._f.values def __call__(self, x): """Evaluates the piecewise linear function at the given point using interpolation""" return self._f(x)
[docs] class piecewise_convex(TransformedPiecewiseLinearFunction): """Simple convex piecewise representation Expresses a piecewise linear function with a convex feasible region for the output variable using a simple collection of linear constraints. """
[docs] def __init__(self, *args, **kwds): super(piecewise_convex, self).__init__(*args, **kwds) breakpoints = self.breakpoints values = self.values self.c = constraint_list() for i in range(len(breakpoints) - 1): X0 = breakpoints[i] F_AT_X0 = values[i] dF_AT_X0 = (values[i + 1] - F_AT_X0) / (breakpoints[i + 1] - X0) const = F_AT_X0 - dF_AT_X0 * X0 con = linear_constraint((self.output, self.input), (-1, dF_AT_X0)) if self.bound == 'ub': con.lb = -const elif self.bound == 'lb': con.ub = -const else: assert self.bound == 'eq' con.rhs = -const self.c.append(con) # In order to enforce the same behavior as actual # piecewise constraints, we need to constrain the # input expression to be between first and last # breakpoint. This might be duplicating the # variable, but its not always the case, and there's # no guarantee that the input "variable" is not a # more general linear expression. self.c.append( linear_constraint( terms=[(self.input, 1)], lb=self.breakpoints[0], ub=self.breakpoints[-1] ) )
[docs] def validate(self, **kwds): """ Validate this piecewise linear function by verifying various properties of the breakpoints, values, and input variable (e.g., that the list of breakpoints is nondecreasing). See base class documentation for keyword descriptions. """ ftype = super(piecewise_convex, self).validate(**kwds) if (self.bound == 'eq') and (ftype != characterize_function.affine): raise PiecewiseValidationError( "The bound type is 'eq' but the function " "was not characterized as affine (only two " "breakpoints). The 'convex' piecewise " "representation does not support this function." ) elif (self.bound == 'lb') and ( ftype not in (characterize_function.affine, characterize_function.convex) ): raise PiecewiseValidationError( "The bound type is 'lb' but the function " "was not characterized as convex or affine. " "The 'convex' piecewise representation does " "not support this function." ) elif (self.bound == 'ub') and ( ftype not in (characterize_function.affine, characterize_function.concave) ): raise PiecewiseValidationError( "The bound type is 'ub' but the function " "was not characterized as concave or affine. " "The 'convex' piecewise representation does " "not support this function." ) return ftype
registered_transforms['convex'] = piecewise_convex
[docs] class piecewise_sos2(TransformedPiecewiseLinearFunction): """Discrete SOS2 piecewise representation Expresses a piecewise linear function using the SOS2 formulation. """
[docs] def __init__(self, *args, **kwds): super(piecewise_sos2, self).__init__(*args, **kwds) # create vars y_tuple = tuple(variable(lb=0) for i in range(len(self.breakpoints))) y = self.v = variable_tuple(y_tuple) # create piecewise constraints self.c = constraint_list() self.c.append( linear_constraint( variables=y_tuple + (self.input,), coefficients=self.breakpoints + (-1,), rhs=0, ) ) self.c.append( linear_constraint( variables=y_tuple + (self.output,), coefficients=self.values + (-1,) ) ) if self.bound == 'ub': self.c[-1].lb = 0 elif self.bound == 'lb': self.c[-1].ub = 0 else: assert self.bound == 'eq' self.c[-1].rhs = 0 self.c.append( linear_constraint(variables=y_tuple, coefficients=(1,) * len(y), rhs=1) ) self.s = sos2(y)
[docs] def validate(self, **kwds): """ Validate this piecewise linear function by verifying various properties of the breakpoints, values, and input variable (e.g., that the list of breakpoints is nondecreasing). See base class documentation for keyword descriptions. """ return super(piecewise_sos2, self).validate(**kwds)
registered_transforms['sos2'] = piecewise_sos2
[docs] class piecewise_dcc(TransformedPiecewiseLinearFunction): """Discrete DCC piecewise representation Expresses a piecewise linear function using the DCC formulation. """
[docs] def __init__(self, *args, **kwds): super(piecewise_dcc, self).__init__(*args, **kwds) # create index sets polytopes = range(len(self.breakpoints) - 1) vertices = range(len(self.breakpoints)) def polytope_verts(p): return range(p, p + 2) # create vars self.v = variable_dict() lmbda = self.v['lambda'] = variable_dict( ((p, v), variable(lb=0)) for p in polytopes for v in vertices ) y = self.v['y'] = variable_tuple( variable(domain_type=IntegerSet, lb=0, ub=1) for p in polytopes ) # create piecewise constraints self.c = constraint_list() self.c.append( linear_constraint( variables=tuple( lmbda[p, v] for p in polytopes for v in polytope_verts(p) ) + (self.input,), coefficients=tuple( self.breakpoints[v] for p in polytopes for v in polytope_verts(p) ) + (-1,), rhs=0, ) ) self.c.append( linear_constraint( variables=tuple( lmbda[p, v] for p in polytopes for v in polytope_verts(p) ) + (self.output,), coefficients=tuple( self.values[v] for p in polytopes for v in polytope_verts(p) ) + (-1,), ) ) if self.bound == 'ub': self.c[-1].lb = 0 elif self.bound == 'lb': self.c[-1].ub = 0 else: assert self.bound == 'eq' self.c[-1].rhs = 0 clist = [] for p in polytopes: variables = tuple(lmbda[p, v] for v in polytope_verts(p)) clist.append( linear_constraint( variables=variables + (y[p],), coefficients=(1,) * len(variables) + (-1,), rhs=0, ) ) self.c.append(constraint_tuple(clist)) self.c.append( linear_constraint(variables=tuple(y), coefficients=(1,) * len(y), rhs=1) )
[docs] def validate(self, **kwds): """ Validate this piecewise linear function by verifying various properties of the breakpoints, values, and input variable (e.g., that the list of breakpoints is nondecreasing). See base class documentation for keyword descriptions. """ return super(piecewise_dcc, self).validate(**kwds)
registered_transforms['dcc'] = piecewise_dcc
[docs] class piecewise_cc(TransformedPiecewiseLinearFunction): """Discrete CC piecewise representation Expresses a piecewise linear function using the CC formulation. """
[docs] def __init__(self, *args, **kwds): super(piecewise_cc, self).__init__(*args, **kwds) # create index sets polytopes = range(len(self.breakpoints) - 1) vertices = range(len(self.breakpoints)) def vertex_polys(v): if v == 0: return [v] if v == len(self.breakpoints) - 1: return [v - 1] else: return [v - 1, v] # create vars self.v = variable_dict() lmbda = self.v['lambda'] = variable_tuple(variable(lb=0) for v in vertices) y = self.v['y'] = variable_tuple( variable(domain_type=IntegerSet, lb=0, ub=1) for p in polytopes ) lmbda_tuple = tuple(lmbda) # create piecewise constraints self.c = constraint_list() self.c.append( linear_constraint( variables=lmbda_tuple + (self.input,), coefficients=self.breakpoints + (-1,), rhs=0, ) ) self.c.append( linear_constraint( variables=lmbda_tuple + (self.output,), coefficients=self.values + (-1,) ) ) if self.bound == 'ub': self.c[-1].lb = 0 elif self.bound == 'lb': self.c[-1].ub = 0 else: assert self.bound == 'eq' self.c[-1].rhs = 0 self.c.append( linear_constraint( variables=lmbda_tuple, coefficients=(1,) * len(lmbda), rhs=1 ) ) clist = [] for v in vertices: variables = tuple(y[p] for p in vertex_polys(v)) clist.append( linear_constraint( variables=variables + (lmbda[v],), coefficients=(1,) * len(variables) + (-1,), lb=0, ) ) self.c.append(constraint_tuple(clist)) self.c.append( linear_constraint(variables=tuple(y), coefficients=(1,) * len(y), rhs=1) )
[docs] def validate(self, **kwds): """ Validate this piecewise linear function by verifying various properties of the breakpoints, values, and input variable (e.g., that the list of breakpoints is nondecreasing). See base class documentation for keyword descriptions. """ return super(piecewise_cc, self).validate(**kwds)
registered_transforms['cc'] = piecewise_cc
[docs] class piecewise_mc(TransformedPiecewiseLinearFunction): """Discrete MC piecewise representation Expresses a piecewise linear function using the MC formulation. """
[docs] def __init__(self, *args, **kwds): super(piecewise_mc, self).__init__(*args, **kwds) # create indexers polytopes = range(len(self.breakpoints) - 1) # create constants (using future division) # these might also be expressions if the breakpoints # or values lists contain mutable objects slopes = tuple( (self.values[p + 1] - self.values[p]) / (self.breakpoints[p + 1] - self.breakpoints[p]) for p in polytopes ) intercepts = tuple( self.values[p] - (slopes[p] * self.breakpoints[p]) for p in polytopes ) # create vars self.v = variable_dict() lmbda = self.v['lambda'] = variable_tuple(variable() for p in polytopes) lmbda_tuple = tuple(lmbda) y = self.v['y'] = variable_tuple( variable(domain_type=IntegerSet, lb=0, ub=1) for p in polytopes ) y_tuple = tuple(y) # create piecewise constraints self.c = constraint_list() self.c.append( linear_constraint( variables=lmbda_tuple + (self.input,), coefficients=(1,) * len(lmbda) + (-1,), rhs=0, ) ) self.c.append( linear_constraint( variables=lmbda_tuple + y_tuple + (self.output,), coefficients=slopes + intercepts + (-1,), ) ) if self.bound == 'ub': self.c[-1].lb = 0 elif self.bound == 'lb': self.c[-1].ub = 0 else: assert self.bound == 'eq' self.c[-1].rhs = 0 clist1 = [] clist2 = [] for p in polytopes: clist1.append( linear_constraint( variables=(y[p], lmbda[p]), coefficients=(self.breakpoints[p], -1), ub=0, ) ) clist2.append( linear_constraint( variables=(lmbda[p], y[p]), coefficients=(1, -self.breakpoints[p + 1]), ub=0, ) ) self.c.append(constraint_tuple(clist1)) self.c.append(constraint_tuple(clist2)) self.c.append( linear_constraint(variables=y_tuple, coefficients=(1,) * len(y), rhs=1) )
[docs] def validate(self, **kwds): """ Validate this piecewise linear function by verifying various properties of the breakpoints, values, and input variable (e.g., that the list of breakpoints is nondecreasing). See base class documentation for keyword descriptions. """ ftype = super(piecewise_mc, self).validate(**kwds) # this representation does not support step functions if ftype == characterize_function.step: raise PiecewiseValidationError( "The 'mc' piecewise representation does not support step functions." ) return ftype
registered_transforms['mc'] = piecewise_mc
[docs] class piecewise_inc(TransformedPiecewiseLinearFunction): """Discrete INC piecewise representation Expresses a piecewise linear function using the INC formulation. """
[docs] def __init__(self, *args, **kwds): super(piecewise_inc, self).__init__(*args, **kwds) # create indexers polytopes = range(len(self.breakpoints) - 1) # create vars self.v = variable_dict() delta = self.v['delta'] = variable_tuple(variable() for p in polytopes) delta[0].ub = 1 delta[-1].lb = 0 delta_tuple = tuple(delta) y = self.v['y'] = variable_tuple( variable(domain_type=IntegerSet, lb=0, ub=1) for p in polytopes[:-1] ) # create piecewise constraints self.c = constraint_list() self.c.append( linear_constraint( variables=(self.input,) + delta_tuple, coefficients=(-1,) + tuple( self.breakpoints[p + 1] - self.breakpoints[p] for p in polytopes ), rhs=-self.breakpoints[0], ) ) self.c.append( linear_constraint( variables=(self.output,) + delta_tuple, coefficients=(-1,) + tuple(self.values[p + 1] - self.values[p] for p in polytopes), ) ) if self.bound == 'ub': self.c[-1].lb = -self.values[0] elif self.bound == 'lb': self.c[-1].ub = -self.values[0] else: assert self.bound == 'eq' self.c[-1].rhs = -self.values[0] clist1 = [] clist2 = [] for p in polytopes[:-1]: clist1.append( linear_constraint( variables=(delta[p + 1], y[p]), coefficients=(1, -1), ub=0 ) ) clist2.append( linear_constraint( variables=(y[p], delta[p]), coefficients=(1, -1), ub=0 ) ) self.c.append(constraint_tuple(clist1)) self.c.append(constraint_tuple(clist2))
[docs] def validate(self, **kwds): """ Validate this piecewise linear function by verifying various properties of the breakpoints, values, and input variable (e.g., that the list of breakpoints is nondecreasing). See base class documentation for keyword descriptions. """ return super(piecewise_inc, self).validate(**kwds)
registered_transforms['inc'] = piecewise_inc
[docs] class piecewise_dlog(TransformedPiecewiseLinearFunction): """Discrete DLOG piecewise representation Expresses a piecewise linear function using the DLOG formulation. This formulation uses logarithmic number of discrete variables in terms of number of breakpoints. """
[docs] def __init__(self, *args, **kwds): super(piecewise_dlog, self).__init__(*args, **kwds) breakpoints = self.breakpoints values = self.values if not is_positive_power_of_two(len(breakpoints) - 1): raise ValueError( "The list of breakpoints must be " "of length (2^n)+1 for some positive " "integer n. Invalid length: %s" % (len(breakpoints)) ) # create branching schemes L = log2floor(len(breakpoints) - 1) assert 2**L == len(breakpoints) - 1 B_LEFT, B_RIGHT = self._branching_scheme(L) # create indexers polytopes = range(len(breakpoints) - 1) vertices = range(len(breakpoints)) def polytope_verts(p): return range(p, p + 2) # create vars self.v = variable_dict() lmbda = self.v['lambda'] = variable_dict( ((p, v), variable(lb=0)) for p in polytopes for v in polytope_verts(p) ) y = self.v['y'] = variable_tuple( variable(domain_type=IntegerSet, lb=0, ub=1) for i in range(L) ) # create piecewise constraints self.c = constraint_list() self.c.append( linear_constraint( variables=(self.input,) + tuple(lmbda[p, v] for p in polytopes for v in polytope_verts(p)), coefficients=(-1,) + tuple(breakpoints[v] for p in polytopes for v in polytope_verts(p)), rhs=0, ) ) self.c.append( linear_constraint( variables=(self.output,) + tuple(lmbda[p, v] for p in polytopes for v in polytope_verts(p)), coefficients=(-1,) + tuple(values[v] for p in polytopes for v in polytope_verts(p)), ) ) if self.bound == 'ub': self.c[-1].lb = 0 elif self.bound == 'lb': self.c[-1].ub = 0 else: assert self.bound == 'eq' self.c[-1].rhs = 0 self.c.append( linear_constraint( variables=tuple(lmbda.values()), coefficients=(1,) * len(lmbda), rhs=1 ) ) clist = [] for i in range(L): variables = tuple(lmbda[p, v] for p in B_LEFT[i] for v in polytope_verts(p)) clist.append( linear_constraint( variables=variables + (y[i],), coefficients=(1,) * len(variables) + (-1,), ub=0, ) ) self.c.append(constraint_tuple(clist)) del clist clist = [] for i in range(L): variables = tuple( lmbda[p, v] for p in B_RIGHT[i] for v in polytope_verts(p) ) clist.append( linear_constraint( variables=variables + (y[i],), coefficients=(1,) * len(variables) + (1,), ub=1, ) ) self.c.append(constraint_tuple(clist))
def _branching_scheme(self, L): N = 2**L B_LEFT = [] for i in range(1, L + 1): start = 1 step = N // (2**i) tmp = [] while start < N: tmp.extend(j - 1 for j in range(start, start + step)) start += 2 * step B_LEFT.append(tmp) biglist = range(N) B_RIGHT = [] for i in range(len(B_LEFT)): tmp = [] for j in biglist: if j not in B_LEFT[i]: tmp.append(j) B_RIGHT.append(sorted(tmp)) return B_LEFT, B_RIGHT
[docs] def validate(self, **kwds): """ Validate this piecewise linear function by verifying various properties of the breakpoints, values, and input variable (e.g., that the list of breakpoints is nondecreasing). See base class documentation for keyword descriptions. """ return super(piecewise_dlog, self).validate(**kwds)
registered_transforms['dlog'] = piecewise_dlog
[docs] class piecewise_log(TransformedPiecewiseLinearFunction): """Discrete LOG piecewise representation Expresses a piecewise linear function using the LOG formulation. This formulation uses logarithmic number of discrete variables in terms of number of breakpoints. """
[docs] def __init__(self, *args, **kwds): super(piecewise_log, self).__init__(*args, **kwds) breakpoints = self.breakpoints values = self.values if not is_positive_power_of_two(len(breakpoints) - 1): raise ValueError( "The list of breakpoints must be " "of length (2^n)+1 for some positive " "integer n. Invalid length: %s" % (len(breakpoints)) ) # create branching schemes L = log2floor(len(breakpoints) - 1) S, B_LEFT, B_RIGHT = self._branching_scheme(L) # create indexers polytopes = range(len(breakpoints) - 1) vertices = range(len(breakpoints)) # create vars self.v = variable_dict() lmbda = self.v['lambda'] = variable_tuple(variable(lb=0) for v in vertices) y = self.v['y'] = variable_list( variable(domain_type=IntegerSet, lb=0, ub=1) for s in S ) # create piecewise constraints self.c = constraint_list() self.c.append( linear_constraint( variables=(self.input,) + tuple(lmbda), coefficients=(-1,) + breakpoints, rhs=0, ) ) self.c.append( linear_constraint( variables=(self.output,) + tuple(lmbda), coefficients=(-1,) + values ) ) if self.bound == 'ub': self.c[-1].lb = 0 elif self.bound == 'lb': self.c[-1].ub = 0 else: assert self.bound == 'eq' self.c[-1].rhs = 0 self.c.append( linear_constraint( variables=tuple(lmbda), coefficients=(1,) * len(lmbda), rhs=1 ) ) clist = [] for s in S: variables = tuple(lmbda[v] for v in B_LEFT[s]) clist.append( linear_constraint( variables=variables + (y[s],), coefficients=(1,) * len(variables) + (-1,), ub=0, ) ) self.c.append(constraint_tuple(clist)) del clist clist = [] for s in S: variables = tuple(lmbda[v] for v in B_RIGHT[s]) clist.append( linear_constraint( variables=variables + (y[s],), coefficients=(1,) * len(variables) + (1,), ub=1, ) ) self.c.append(constraint_tuple(clist))
def _branching_scheme(self, n): N = 2**n S = range(n) G = generate_gray_code(n) L = tuple( [ k for k in range(N + 1) if ((k == 0) or (G[k - 1][s] == 1)) and ((k == N) or (G[k][s] == 1)) ] for s in S ) R = tuple( [ k for k in range(N + 1) if ((k == 0) or (G[k - 1][s] == 0)) and ((k == N) or (G[k][s] == 0)) ] for s in S ) return S, L, R
[docs] def validate(self, **kwds): """ Validate this piecewise linear function by verifying various properties of the breakpoints, values, and input variable (e.g., that the list of breakpoints is nondecreasing). See base class documentation for keyword descriptions. """ return super(piecewise_log, self).validate(**kwds)
registered_transforms['log'] = piecewise_log