-
Notifications
You must be signed in to change notification settings - Fork 40
/
sgp.py
248 lines (234 loc) · 11.6 KB
/
sgp.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
"""Implement the SequentialGeometricProgram class"""
import warnings as pywarnings
from time import time
from collections import defaultdict
import numpy as np
from ..exceptions import (InvalidGPConstraint, Infeasible, UnnecessarySGP,
InvalidPosynomial, InvalidSGPConstraint)
from ..keydict import KeyDict
from ..nomials import Variable
from .gp import GeometricProgram
from ..nomials import PosynomialInequality, Posynomial
from .. import NamedVariables
EPS = 1e-6 # 1 +/- this is used in a few relative differences
# pylint: disable=too-many-instance-attributes
class SequentialGeometricProgram:
"""Prepares a collection of signomials for a SP solve.
Arguments
---------
cost : Posynomial
Objective to minimize when solving
constraints : list of Constraint or SignomialConstraint objects
Constraints to maintain when solving (implicitly Signomials <= 1)
verbosity : int (optional)
Currently has no effect: SequentialGeometricPrograms don't know
anything new after being created, unlike GeometricPrograms.
Attributes with side effects
----------------------------
`gps` is set during a solve
`result` is set at the end of a solve
Examples
--------
>>> gp = gpkit.geometric_program.SequentialGeometricProgram(
# minimize
x,
[ # subject to
1/x - y/x, # <= 1, implicitly
y/10 # <= 1
])
>>> gp.solve()
"""
gps = solver_outs = _results = result = model = None
with NamedVariables("SGP"):
slack = Variable("PCCPslack")
def __init__(self, cost, model, substitutions, *,
use_pccp=True, pccp_penalty=2e2, checkbounds=True):
self.pccp_penalty = pccp_penalty
if cost.any_nonpositive_cs:
raise InvalidPosynomial("""an SGP's cost must be Posynomial
The equivalent of a Signomial objective can be constructed by constraining
a dummy variable `z` to be greater than the desired Signomial objective `s`
(z >= s) and then minimizing that dummy variable.""")
self.gpconstraints, self.sgpconstraints = [], []
if not use_pccp:
self.slack = 1
else:
self.gpconstraints.append(self.slack >= 1)
cost *= self.slack**pccp_penalty
self.approxconstraints = []
self.sgpvks = set()
x0 = KeyDict(substitutions)
x0.varkeys = model.varkeys # for string access and so forth
for cs in model.flat():
try:
if not hasattr(cs, "as_hmapslt1"):
raise InvalidGPConstraint(cs)
if not isinstance(cs, PosynomialInequality):
cs.as_hmapslt1(substitutions) # gp-compatible?
self.gpconstraints.append(cs)
except InvalidGPConstraint:
if not hasattr(cs, "as_gpconstr"):
raise InvalidSGPConstraint(cs)
self.sgpconstraints.append(cs)
for hmaplt1 in cs.as_gpconstr(x0).as_hmapslt1({}):
constraint = (Posynomial(hmaplt1) <= self.slack)
constraint.generated_by = cs
self.approxconstraints.append(constraint)
self.sgpvks.update(constraint.varkeys)
if not self.sgpconstraints:
raise UnnecessarySGP("""Model valid as a Geometric Program.
SequentialGeometricPrograms should only be created with Models containing
Signomial Constraints, since Models without Signomials have global
solutions and can be solved with 'Model.solve()'.""")
self._gp = GeometricProgram(
cost, self.approxconstraints + self.gpconstraints,
substitutions, checkbounds=checkbounds)
self._gp.x0 = x0
self.a_idxs = defaultdict(list)
cost_mons = self._gp.k[0]
sp_mons = sum(self._gp.k[:1+len(self.approxconstraints)])
for row_idx, m_idx in enumerate(self._gp.A.row):
if cost_mons <= m_idx <= sp_mons:
self.a_idxs[self._gp.p_idxs[m_idx]].append(row_idx)
# pylint: disable=too-many-locals,too-many-branches,too-many-statements
def localsolve(self, solver=None, *, verbosity=1, x0=None, reltol=1e-4,
iteration_limit=50, **solveargs):
"""Locally solves a SequentialGeometricProgram and returns the solution.
Arguments
---------
solver : str or function (optional)
By default uses one of the solvers found during installation.
If set to "mosek", "mosek_cli", or "cvxopt", uses that solver.
If set to a function, passes that function cs, A, p_idxs, and k.
verbosity : int (optional)
If greater than 0, prints solve time and number of iterations.
Each GP is created and solved with verbosity one less than this, so
if greater than 1, prints solver name and time for each GP.
x0 : dict (optional)
Initial location to approximate signomials about.
reltol : float
Iteration ends when this is greater than the distance between two
consecutive solve's objective values.
iteration_limit : int
Maximum GP iterations allowed.
mutategp: boolean
Prescribes whether to mutate the previously generated GP
or to create a new GP with every solve.
**solveargs :
Passed to solver function.
Returns
-------
result : dict
A dictionary containing the translated solver result.
"""
self.gps, self.solver_outs, self._results = [], [], []
starttime = time()
if verbosity > 0:
print("Starting a sequence of GP solves")
print(" for %i free variables" % len(self.sgpvks))
print(" in %i locally-GP constraints" % len(self.sgpconstraints))
print(" and for %i free variables" % len(self._gp.varlocs))
print(" in %i posynomial inequalities." % len(self._gp.k))
prevcost, cost, rel_improvement = None, None, None
while rel_improvement is None or rel_improvement > reltol:
prevcost = cost
if len(self.gps) > iteration_limit:
raise Infeasible(
"Unsolved after %s iterations. Check `m.program.results`;"
" if they're converging, try `.localsolve(...,"
" iteration_limit=NEWLIMIT)`." % len(self.gps))
gp = self.gp(x0, cleanx0=True)
self.gps.append(gp) # NOTE: SIDE EFFECTS
if verbosity > 1:
print("\nGP Solve %i" % len(self.gps))
if verbosity > 2:
print("===============")
solver_out = gp.solve(solver, verbosity=verbosity-1,
gen_result=False, **solveargs)
self.solver_outs.append(solver_out)
cost = float(solver_out["objective"])
x0 = dict(zip(gp.varlocs, np.exp(solver_out["primal"])))
if verbosity > 2:
result = gp.generate_result(solver_out, verbosity=verbosity-3)
self._results.append(result)
print(result.table(self.sgpvks))
elif verbosity > 1:
print("Solved cost was %.4g." % cost)
if prevcost is None:
continue
rel_improvement = (prevcost - cost)/(prevcost + cost)
if cost*(1 - EPS) > prevcost + EPS and verbosity > -1:
pywarnings.warn("SGP not convergent:"
" Cost rose by %.2g%% on GP solve %i."
" Details can be found in `m.program.results` or by"
" solving at a higher verbosity. Note that convergence"
" is not guaranteed for models with SignomialEqualities."
% (100*(cost - prevcost)/prevcost, len(self.gps)))
rel_improvement = cost = None
# solved successfully!
self.result = gp.generate_result(solver_out, verbosity=verbosity-3)
self.result["soltime"] = time() - starttime
if verbosity > 1:
print()
if verbosity > 0:
print("Solving took %.3g seconds and %i GP solves."
% (self.result["soltime"], len(self.gps)))
if hasattr(self.slack, "key"):
excess_slack = self.result["variables"][self.slack.key] - 1 # pylint: disable=no-member
if excess_slack <= EPS:
del self.result["freevariables"][self.slack.key] # pylint: disable=no-member
del self.result["variables"][self.slack.key] # pylint: disable=no-member
del self.result["sensitivities"]["variables"][self.slack.key] # pylint: disable=no-member
slackconstraint = self.gpconstraints[0]
del self.result["sensitivities"]["constraints"][slackconstraint]
elif verbosity > -1:
print("Final solution let signomial constraints slacken by"
" %.2g%%. Calling .localsolve with a higher"
" `pccp_penalty` (it was %.3g this time) will reduce"
" final slack if the model is solvable with less. If"
" you think it might not be, check by solving with "
"`use_pccp=False, x0=(this model's final solution)`.\n"
% (100*excess_slack, self.pccp_penalty))
return self.result
@property
def results(self):
"Creates and caches results from the raw solver_outs"
if not self._results:
self._results = [o["generate_result"]() for o in self.solver_outs]
return self._results
def gp(self, x0=None, *, cleanx0=False):
"Update self._gp for x0 and return it."
if not x0:
return self._gp # return last generated
if not cleanx0:
x0 = KeyDict(x0)
self._gp.x0.update({vk: x0[vk] for vk in self.sgpvks if vk in x0})
p_idx = 0
for sgpc in self.sgpconstraints:
for hmaplt1 in sgpc.as_gpconstr(self._gp.x0).as_hmapslt1({}):
approxc = self.approxconstraints[p_idx]
approxc.unsubbed = [Posynomial(hmaplt1)/self.slack]
p_idx += 1 # p_idx=0 is the cost; sp constraints are after it
hmap, = approxc.as_hmapslt1(self._gp.substitutions)
self._gp.hmaps[p_idx] = hmap
m_idx = self._gp.m_idxs[p_idx].start
a_idxs = list(self.a_idxs[p_idx]) # A's entries we can modify
for i, (exp, c) in enumerate(hmap.items()):
self._gp.exps[m_idx + i] = exp
self._gp.cs[m_idx + i] = c
for var, x in exp.items():
try: # modify a particular A entry
row_idx = a_idxs.pop()
self._gp.A.row[row_idx] = m_idx + i
self._gp.A.col[row_idx] = self._gp.varidxs[var]
self._gp.A.data[row_idx] = x
except IndexError: # numbers of exps increased
self.a_idxs[p_idx].append(len(self._gp.A.row))
self._gp.A.row.append(m_idx + i)
self._gp.A.col.append(self._gp.varidxs[var])
self._gp.A.data.append(x)
for row_idx in a_idxs: # number of exps decreased
self._gp.A.row[row_idx] = 0 # zero out this entry
self._gp.A.col[row_idx] = 0
self._gp.A.data[row_idx] = 0
return self._gp