How to extract numerical values from the LinExpr objects

Answered

Babooshka Shavazipour

October 10, 2023 07:44

• 
• 

Hi,

I want to formulate an objective function to calculate VaR.

I get errors when I try to run the following code:

# A function to calculate the scnearioGainLoss

def objective_function(x, confidence_interval):

# Monte Carlo simulations

# run 10 000 simulations for the Loss function based on pre-generated scenarios

simulated_losses= []

foriinrange(num_simulations):

# calculate the loss for the current simulation

loss=sum(x[j] *np.array(simulated_value[i])[j] for j in range(len(df))) -D # D is a constant, similar error when using quicksum()

# Append the loss to the list of simulated losses

simulated_losses.append(loss)

beta=1-confidence_interval

 

# Sort the simulated losses

simulated_losses.sort()

#sorted_simulated_losses = sorted(simulated_losses) # the same error

 

# Calculate VaR

var_index=int(beta*num_simulations)

VaR=-simulated_losses[var_index]

# VaR = -np.percentile(scenarioLoss_values, 100 * (1-confidence_interval)) # similar error

return VaR

m = Model("milp")

# Create binary decision variables x_i

x = m.addVars(len(df), lb=0, ub=1, vtype=GRB.BINARY, name="x")

confidence_interval=0.95

# Defining objective function

obj_fun = objective_function(x, confidence_interval)

m.setObjective(obj_fun, GRB.MINIMIZE)

 

---------------------------------------------------------------------------

NotImplementedError             Traceback (most recent call last) Cell In[128], line 33 30 confidence_interval = 0.95 32 # Defining objective function ---> 33 obj_fun = objective_function(x, confidence_interval) 34 m.setObjective(obj_fun, GRB.MINIMIZE) Cell In[128], line 15, in objective_function(x, confidence_interval) 12 beta = 1 - confidence_interval 14 # Sort the simulated losses ---> 15 simulated_losses.sort() 16 #sorted_simulated_losses = sorted(simulated_losses) 17 18 # Calculate VaR 19 var_index = int(beta * num_simulations) File src/gurobipy/linexpr.pxi:473, in gurobipy.LinExpr.__richcmp__() NotImplementedError:

 

PS: It seems that the simulated_losses list contains Gurobi LinExpr objects. I do not know how to extract numerical values from the LinExpr objects in this case. I tried 

[expr.getValue() for expr in simulated_losses], but it does not work. 

0

• 

  Marika Karbstein

  • Gurobi Staff

  October 10, 2023 10:09

  You are creating the Gurobi model when defining the list simulated_lossed. That means there is no solution value for the x variables. Mathematical operations on the x-variables give you linear expressions. But getValue() is not possible as long as the model is not solved.

  Hence, the list simulated_lossed contains linear expressions that depend on the model variables x.
  Now, you can define linear constraints on the expressions. In the documentation Constraints - Gurobi Optimization you can see which type of constraints are possible and can be added to a Gurobi model.
  You can define for example min and max constraints (see also Model.addGenConstrMin() and Model.addGenConstrMax() ). But there is no such thing as a sorting constraint.
  So what you try to achieve is not possible.
  Try to formulate your problem with variables and constraints based on the documentation Variables and Constraints and Objectives.

  0

Please sign in to leave a comment.