mle-decoder

Most-likely error (MLE) decoder that uses gurobi to solve the mixed-interger (linear) programming problem.

The decoding hypergraph is specified using stim.DetectorErrorModel.

MLE optimization description

The problem of finding the most likely error can be mapped to a mixed-integer (linear) programming (MILP) problem. This can be solved by optimization solvers like Gurobi. The mapping to a MILP problem is (source: https://arxiv.org/abs/2403.03272)

Note that $C_i \in \{-1, +1\}$.

Setting up the gurobi license

1. Create a free academic account
2. Request a license, which will give you a license key
3. Install the Gurobi Optimizer (or install gurobipy through conda) so that we can run the grbgetkey command
4. Run grbgetkey xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx using your license number
5. At the end of ~/.bashrc add export GRB_LICENSE_FILE=/path/to/license.lic where the license path is printed when running the previous step
6. Run source ~/.bashrc or open a new terminal and check that the license installation is successful by running the gurobi.sh command