matrix modification

[LeonidPryadko]

Implement workflow for matrix modification for a quantum CSS code (star-polygon transformation in the equivalent Ising model).

• read in codewords file
• eliminate rows of G matrix of weight 1, 2 (inverse decoration transformation), and 3 (star-triangle transformation)
• eliminate rows of general weight
• ensure P vectors can be both read and written to a file
• verify LLR transformation formulas
• add general information with debug=2 about DEM or matrices read (dimensions, ranks, redundant rows, row and column weight, etc.)
• maxW command line argument for a hard upper bound of codewords to process.
• coordinate and syndrome map between the original and modified code; also for multi-step modification (introduce a transformation matrix T -- is one such matrix sufficient?)
• come up with the protocol to use modified matrices with errors sampled internally.
• verify transformation using old Mathematica code and Ising partition functions
• come up with the protocol to use modified matrices with errors sampled externally (e.g., in the original code using the T matrix or without).
• Come up with the replacement for 'pads' command line argument to enable using the same det and obs events
• Why is the decoding accuracy degraded so badly? When ML or near-ML decoder is available, try with such a decoder -- should have the same success probability