Making optimization simple (Python)

Optimization (Prescriptive analytics, Operations Research, Decision Optimization) is doing more with less. Which is why most of the time, this makes sense. In a few posts I tried to show the way:

• Low barrier to entry to Optimization
• Should we use a dedicated modeling language or a general purpose language ?

Then came 2018 and I tried a very simple story.

And many people liked Making Decision Optimization Simple where I gave some very simple OPL examples derived from the very simple bus example.

Python is everywhere especially thanks to the data science boom and to me the easiest path in order to use decision optimization in python is to either use the OPL - Python doopl API or to call OPL in Watson Studio either local or cloud.

Some people do not answer the same as me to the question whether we should use OPL from Python or Python directly and in this post, I d like to please them.

A few docplex python examples:

• basic zoo example
• with Constraint Programming
• Get solution information for MP and CP
• scheduling

• data in a list (v2,v3)
• the same but with constraint programming (v2,v3)
• data in a dataframe (v2,v3)
• data in a class
• function in an objective
• incremental changes
• several scenarii in a pandas data frame

• in a Watson Studio notebook
• range constraints
• logical constraints
• max constraint
• absolute value
• disjunctions (or) and with Constraint Programming
• if then (v2) (if static condition)
• quadratic
• SOS1 and SOS2 constraints
• all different
• all and any
• lazy constraint
• black-box constraint

• modulo constraint with cplex and constraint programming
• decision variable in a domain with cplex and cpo
• with some cplex settings
• decimal variables with CPO
• get cplex info after solve
• get the gap
• save solution into a file
• relaxation and conflicts
• hard and soft constraints
• relax integrity and find dual
• relax one integer decision variable
• Solvefixed, dual values and reduced cost
• piecewise linear function
• non linear functions
• table function

• solution pools
• get all optimal solutions
• enumerate without solution pools
• enumerate solutions within constraint programming
• get and set coef
• kpi with linear model and constraint programming
• warm start ( write / read )
• warm start through API
• warmstart within constraint programming
• Fixed start with MP and constraint programming
• clone a linear model and constraint programming
• monte carlo simulation
• monte carlo optimization (which you can also run in parallel)
• robust optimization
• stochastic optimization
• multiobjective and with CPOptimizer
• semiinteger and semicontinuous
• count true constraints
• count unique values MP and Constraint Programming
• set branch priorities MIP and CP
• progress listener / MIP info callback
• get solutions one by one
• export LP format

and we can call OPL model from python and

• call an OPL model from python and then add python objective and constraints (MIP , CPO, and scheduling )

There are many CPLEX Python API but the one I recommend is docplex. Good documentation for Mathematical Programming and Constraint Programming.

There are many available tutorial notebooks. And a very convenient way to run docplex python code is to use IBM Watson Studio.

Plus with docplex you can solve with IBMQuantum through an emulator or a real quantum device

PS:

Let's not forget since we deal with Python that all started with some excellent movies and some nonsense:

• Knights
• Witches