def updateObjFcnAndSolve(self, score_weights, pref_weight, e_cap_weight,
congestion_weight, dept_fairness, b2b_weight):
"""choiceweights should be in order [1st choice, 2nd choice, etc]"""
pref_weight = max(pref_weight, self.config.EPS_SAFETY_OVERRIDE)
#normalize the score weights
max_weight = float(max(score_weights))
if max_weight == 0:
max_weight = 1.
score_weights = [w/max_weight for w in score_weights]
#check to see if fairness constraints are already there
if self.hasDeptFairness:
const_names = ["DeptFairness_%s" % dept for dept in self.getDepts()]
self.m.linear_constraints.delete(const_names)
self.addDeptFairnessConstraints(score_weights)
#normalize weights to make comparable
e_cap_weight /= float(-len(self.course_list))
pref_weight /= float(len(self.course_list))
#VG better performance if we don't normalize here...
#b2b_weight /= float(len(self.b2b_vars) + 1 ) #add 1 for safety
obj_coefs = []
for c, r, t, var in self.vars:
#ecap weight
coef_ecap = helpers.e_cap(c, r) * e_cap_weight
#preference business
coef_pref = 0
for ix in range(len(c.roomPrefs)):
if c.roomPrefs[ix] == r:
coef_pref += score_weights[ix]
for ix in range(len(c.timePrefs)):
if c.timePrefs[ix] == t:
coef_pref += score_weights[ix]
#add large penalty if day string doesn't match prefferred
coef_pref_day = 0.
if not c.isPreferredDays(t):
coef_pref_day = -self.config.SOFT_CNST_PENALTY
obj_coefs.append(coef_pref * pref_weight + coef_ecap + coef_pref_day)
self.m.objective.set_sense(self.m.objective.sense.maximize)
#add the maxCong
vars_only = [v for c, r, t, v in self.vars]
vars_only += [self.maxCongVar]
obj_coefs += [-congestion_weight]
#add minDeptFairness
vars_only += [self.minDept]
obj_coefs += [dept_fairness]
#add the bonuses for b2b teaching
vars_only += self.b2b_vars
obj_coefs += [b2b_weight] * len(self.b2b_vars)
self.m.objective.set_linear(zip(vars_only, obj_coefs))
self.m.solve()
solution = self.m.solution
#Debug
#print solution.get_status(), solution.status[solution.get_status()],
if solution.get_status() not in (101, 102): #fix this
sol_status = solution.get_status()
#probably infeasible
self.m.conflict.refine(self.m.conflict.all_constraints())
self.m.conflict.write("infeasible_conflict")
self.m.write("ses.lp")
print "VG Sol_status", sol_status
print "VG solution string", solution.status[sol_status]
raise ses.SESError("Optimizer did not solve. Check ses.lp Status and infeasible_conflict: %s %s" %
(solution.status[sol_status], sol_status) )
def updateObjFcnAndSolve(self, score_weights, pref_weight, e_cap_weight,
congestion_weight, dept_fairness, b2b_weight):
"""choiceweights should be in order [1st choice, 2nd choice, etc]"""
pref_weight = max(pref_weight, self.config.EPS_SAFETY_OVERRIDE)
#normalize the score weights
max_weight = float(max(score_weights))
if max_weight == 0:
max_weight = 1.
score_weights = [w / max_weight for w in score_weights]
#check to see if fairness constraints are already there
if self.FairnessConstraints:
[self.m.remove(const) for const in self.FairnessConstraints]
self.FairnessConstraints = []
self.addDeptFairnessConstraints(score_weights)
#Normalize weights to make order unity
e_cap_weight /= float(-len(self.course_list))
pref_weight /= float(len(self.course_list))
obj = grb.LinExpr()
for c, r, t, var in self.vars:
obj += helpers.e_cap(c, r) * e_cap_weight * var
#add the preference weights
for c, r, t, var in self.vars:
for ix in range(len(c.roomPrefs)):
if c.roomPrefs[ix] == r:
obj += score_weights[ix] * pref_weight * var
for ix in range(len(c.timePrefs)):
if c.timePrefs[ix] == t:
obj += score_weights[ix] * pref_weight * var
#add large penalty if day string doesn't match prefferred
if not c.isPreferredDays(t):
obj -= self.config.SOFT_CNST_PENALTY * var
#add the maxCong, minDeptFairness
obj += -congestion_weight * self.maxCongVar
obj += dept_fairness * self.minDept
#add the bonuses for b2b teaching
obj += grb.quicksum(b2b_weight * v for v in self.b2b_vars)
self.m.setObjective(obj, grb.GRB.MAXIMIZE)
self.m.optimize()
if self.m.status == grb.GRB.status.INF_OR_UNBD:
self.m.params.presolve = 0
self.m.optimize()
if self.m.status == grb.GRB.status.INFEASIBLE:
self.m.computeIIS()
self.m.write("infeasible_conflict.ilp")
raise ses.SESError(
"Optimization Infeasible. Check file infeasible_conflict.ilp")
elif self.m.status <> grb.GRB.status.OPTIMAL:
self.m.write("ses.lp")
raise SESError("Optimizer did not solve. Check ses.lp Status: %d" %
self.m.status)
def excessCap(self):
"""Compute excess capacity for every course"""
f = lambda c: e_cap(c, c.assignedRoom)
return map(f, self.courses_filt)
def updateObjFcnAndSolve(self, score_weights, pref_weight, e_cap_weight,
congestion_weight, dept_fairness, b2b_weight):
"""choiceweights should be in order [1st choice, 2nd choice, etc]"""
pref_weight = max(pref_weight, self.config.EPS_SAFETY_OVERRIDE)
#normalize the score weights
max_weight = float(max(score_weights))
if max_weight == 0:
max_weight = 1.
score_weights = [w/max_weight for w in score_weights]
#check to see if fairness constraints are already there
if self.FairnessConstraints:
[self.m.remove(const) for const in self.FairnessConstraints]
self.FairnessConstraints = []
self.addDeptFairnessConstraints(score_weights)
#Normalize weights to make order unity
e_cap_weight /= float(-len(self.course_list))
pref_weight /= float(len(self.course_list))
obj = grb.LinExpr()
for c, r, t, var in self.vars:
obj += helpers.e_cap(c, r) * e_cap_weight * var
#add the preference weights
for c, r, t, var in self.vars:
for ix in range(len(c.roomPrefs)):
if c.roomPrefs[ix] == r:
obj += score_weights[ix] * pref_weight * var
for ix in range(len(c.timePrefs)):
if c.timePrefs[ix] == t:
obj += score_weights[ix] * pref_weight * var
#add large penalty if day string doesn't match prefferred
if not c.isPreferredDays(t):
obj -= self.config.SOFT_CNST_PENALTY * var
#add the maxCong, minDeptFairness
obj += -congestion_weight * self.maxCongVar
obj += dept_fairness * self.minDept
#add the bonuses for b2b teaching
obj += grb.quicksum(b2b_weight * v for v in self.b2b_vars)
self.m.setObjective(obj, grb.GRB.MAXIMIZE)
self.m.optimize()
if self.m.status == grb.GRB.status.INF_OR_UNBD:
self.m.params.presolve = 0
self.m.optimize()
if self.m.status == grb.GRB.status.INFEASIBLE:
self.m.computeIIS()
self.m.write("infeasible_conflict.ilp")
raise ses.SESError("Optimization Infeasible. Check file infeasible_conflict.ilp")
elif self.m.status <> grb.GRB.status.OPTIMAL:
self.m.write("ses.lp")
raise SESError("Optimizer did not solve. Check ses.lp Status: %d" % self.m.status)
def summarizeExcessCap(self):
"""Compute Avg Excess capacity"""
f = lambda c: e_cap(c, c.assignedRoom)
return 100 * sum(map(f, self.courses_filt))/len(self.courses_filt)
def updateObjFcnAndSolve(self, score_weights, pref_weight, e_cap_weight,
congestion_weight, dept_fairness, b2b_weight):
"""choiceweights should be in order [1st choice, 2nd choice, etc]"""
pref_weight = max(pref_weight, self.config.EPS_SAFETY_OVERRIDE)
#normalize the score weights
max_weight = float(max(score_weights))
if max_weight == 0:
max_weight = 1.
score_weights = [w / max_weight for w in score_weights]
#check to see if fairness constraints are already there
if self.hasDeptFairness:
const_names = [
"DeptFairness_%s" % dept for dept in self.getDepts()
]
self.m.linear_constraints.delete(const_names)
self.addDeptFairnessConstraints(score_weights)
#normalize weights to make comparable
e_cap_weight /= float(-len(self.course_list))
pref_weight /= float(len(self.course_list))
#VG better performance if we don't normalize here...
#b2b_weight /= float(len(self.b2b_vars) + 1 ) #add 1 for safety
obj_coefs = []
for c, r, t, var in self.vars:
#ecap weight
coef_ecap = helpers.e_cap(c, r) * e_cap_weight
#preference business
coef_pref = 0
for ix in range(len(c.roomPrefs)):
if c.roomPrefs[ix] == r:
coef_pref += score_weights[ix]
for ix in range(len(c.timePrefs)):
if c.timePrefs[ix] == t:
coef_pref += score_weights[ix]
#add large penalty if day string doesn't match prefferred
coef_pref_day = 0.
if not c.isPreferredDays(t):
coef_pref_day = -self.config.SOFT_CNST_PENALTY
obj_coefs.append(coef_pref * pref_weight + coef_ecap +
coef_pref_day)
self.m.objective.set_sense(self.m.objective.sense.maximize)
#add the maxCong
vars_only = [v for c, r, t, v in self.vars]
vars_only += [self.maxCongVar]
obj_coefs += [-congestion_weight]
#add minDeptFairness
vars_only += [self.minDept]
obj_coefs += [dept_fairness]
#add the bonuses for b2b teaching
vars_only += self.b2b_vars
obj_coefs += [b2b_weight] * len(self.b2b_vars)
self.m.objective.set_linear(zip(vars_only, obj_coefs))
self.m.solve()
solution = self.m.solution
#Debug
#print solution.get_status(), solution.status[solution.get_status()],
if solution.get_status() not in (101, 102): #fix this
sol_status = solution.get_status()
#probably infeasible
self.m.conflict.refine(self.m.conflict.all_constraints())
self.m.conflict.write("infeasible_conflict")
self.m.write("ses.lp")
print "VG Sol_status", sol_status
print "VG solution string", solution.status[sol_status]
raise ses.SESError(
"Optimizer did not solve. Check ses.lp Status and infeasible_conflict: %s %s"
% (solution.status[sol_status], sol_status))
def excessCap(self):
"""Compute excess capacity for every course"""
f = lambda c: e_cap(c, c.assignedRoom)
return map(f, self.courses_filt)
def summarizeExcessCap(self):
"""Compute Avg Excess capacity"""
f = lambda c: e_cap(c, c.assignedRoom)
return 100 * sum(map(f, self.courses_filt)) / len(self.courses_filt)
def test_e_cap(self):
self.assertEqual(helpers.e_cap(self.course3, self.room3), 0)
self.assertEqual(helpers.e_cap(self.course3, self.room1), 0.15)