The inference method based on the maximum entropy principle (MaxEnt principle) asserts that the most suitable probability distribution compatible with a given set of constraints is the one with the largest entropy. This method is considered as a powerful estimation technique in a wide range of probabilistic models since it brings a solution to the universal problem of trying to stract information from partial or incomplete data.
Finding the MaxEnt probability distribution is considered a hard task due to the nonlinearities in the reconstruction algorithm. We have redefined the MaxEnt problem as a QUBO problem using a linearized entropy. As a proof of concept, we present a code that finds the MaxEnt probability distribution given the appearance frequencies of the faces of a dice as constraints.
- Make sure you have Ocean installed and configured. Follow the instructions.
- You need to have Jupyter Notebook installed. If you do not have it installed, follow this instructions.
To see the example showed in the code, all you need to do is run the cells pressing Shift+Enter.
We proposed a solution for the DWave challenge.
Inés Corte, Federico Holik, Marcelo Losada, Lorena Rebón, Diego Tielas