add_predictor_constr in Gurobi modeling

Answered

Florencia Vidal

• May 07, 2023 04:12
• Edited

Hi everyone,

I'm currently working on modifying an existing notebook titled "Part Two: Avocado Pricing and Supply Using Mathematical Optimization" available at https://github.com/Gurobi/modeling-examples/blob/master/price_optimization/price_optimization_gurobiML_gcl.ipynb.

My objective is to apply logarithmic transformations to the price and units_sold variables, but I'm facing an issue when trying to use the add_predictor_constr function. Unfortunately, I'm not able to figure out what the issue is, so I would greatly appreciate any assistance or suggestions on how to resolve this issue.

This is the link to my code https://github.com/Florenciavidal/TID/blob/main/log_transform3ipynb.ipynb 

Thank you in advance for your help!

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

UnboundLocalError                         Traceback (most recent call last)

<ipython-input-664-dfaa36de4a08> in <cell line: 1>()
----> 1 pred_constr = add_predictor_constr(m, reg, m_feats, d)
      2 pred_constr.print_stats()

---

13 frames

---

/usr/local/lib/python3.10/dist-packages/gurobi_ml/sklearn/column_transformer.py in _mip_model(self, **kwargs)
     77                     if name in transformers:
     78                         trans_constr = transformers[name](self._gp_model, trans, data)
---> 79                     transformed.append(trans_constr.output.tolist())
     80                 else:
     81                     transformed.append(trans.transform(_input.loc[:, cols]))

UnboundLocalError: local variable 'trans_constr' referenced before assignment

Related to

• gurobi-ml

0

• Official comment

  

  Pierre Bonami

  Gurobi Staff

  • May 09, 2023 11:06

  Hi Florencia,

  The issue is that there is no implementation for adding the logarithmic transformation in Gurobi ML. Since the transformation is applied to Gurobi variables in the model, one would need to implement a class that performs it.

  Then the error was not very graceful that I will fix.

  If you really want to use a logarithmic transformer you would need to implement a small class similar to the ones in https://github.com/Gurobi/gurobi-machinelearning/blob/main/src/gurobi_ml/sklearn/preprocessing.py.

  I hope this helps. Best regards,

  Pierre
• 

  Riley Clement

  Gurobi Staff

  • May 08, 2023 17:38

  Hi Florencia,

  I suspect this may be a bug in gurobi-machinelearning.

  I have created a Github Issue for it.

  - Riley

   

  0

Please sign in to leave a comment.