mdeboute / graph_partition_optim

A graph partition optimization project

Hello world! 👋

I'm Martin, a young engineer with a passion for operations research and optimization for decision support. I'm convinced that my field of expertise can make a significant contribution to improving our society.

🎯 My main aspiration is to promote operational research as an essential tool for making informed decisions in all fields. My commitment is to deploy my skills to contribute to this goal. I'm also interested in other related disciplines, for me continuing education is necessary to stay relevant in this ever-changing world.

🌍 Outside of work, you'll often find me playing sports. I firmly believe that the balance between the physical and the intellectual is essential. So whether it's solving equations or playing sport, I'm always up for a new challenge!

🤝 I'm open to new collaborations and exploring innovative projects. So if you share my vision and believe as I do in the potential of optimization to create a positive impact in this world, please don't hesitate to contact me.

🔮 Tools, Technologies and Languages

graph_partition_optim's Issues

Enhancement

A visual comparison of the quality of solutions and time between the different approaches was missing.
The 'strict' equality approach is not reasonable for a real problem.
We are facing a multi-objective problem, it is important to think about the relaxable constraints.

Moreover, stochastic algorithms (gradient, annealing, tabu) are sensitive to initialization when comparing these different approaches, we have to keep in mind :

a/ that the comparison must be done with the same initialization
b/ that the comparison on average must also concern nb best cases/average/worst case
c/ that the time limit must be related to the nature of the instances (60s for a pb n=100 cannot be 60s with n=10000)

Without being linear it is necessary to allow a little more flexibility.