A Julia meta-package for accessing the python libraries pymatching
(for MWPM-like decoders) and ldpc
(for BP-like decoders).
The python pymatching module is immediately available:
julia> using PyQDecoders
julia> PyQDecoders.pm
Python: <module 'pymatching' from ...>
Running the example from pymatching
's original Readme:
julia> using PyQDecoders: sps, np, pm
julia> H = sps.csc_matrix(
[1 1 0 0 0;
0 1 1 0 0;
0 0 1 1 0;
0 0 0 1 1])
Python:
<4x5 sparse matrix of type '<class 'numpy.int64'>'
with 8 stored elements in Compressed Sparse Column format>
julia> weights = [4, 3, 2, 3, 4]
5-element Vector{Int64}:
4
3
2
3
4
julia> matching = pm.Matching(H, weights=weights)
Python: <pymatching.Matching object with 4 detectors, 1 boundary node, and 5 edges>
julia> prediction = matching.decode([0, 1, 0, 1])
Python: array([0, 0, 1, 1, 0], dtype=uint8)
julia> prediction, solution_weight = matching.decode([0, 1, 0, 1], return_weight=true);
julia> prediction
Python: array([0, 0, 1, 1, 0], dtype=uint8)
julia> solution_weight
Python: 5.0
The python ldpc module is immediately available:
julia> using PyQDecoders
julia> PyQDecoders.ldpc
Python: <module 'ldpc' from ...>
Running the example from ldpc
's original Readme:
julia> using PyQDecoders: sps, np, ldpc
julia> H = ldpc.codes.rep_code(3) # parity check matrix for the length-3 repetition code
Python:
array([[1, 1, 0],
[0, 1, 1]])
julia> n = H.shape[1] # the codeword length, same as `size(H, 2)`
Python: 3
julia> bpd = ldpc.bp_decoder(
H, # the parity check matrix
error_rate=0.1, # the error rate on each bit
max_iter=n, # the maximum iteration depth for BP
bp_method="product_sum", # BP method. The other option is `minimum_sum'
channel_probs=[nothing] # channel probability probabilities. Will override error rate.
)
Python: <ldpc.bp_decoder.bp_decoder object at 0x...>
julia> error = [0,1,0]
3-element Vector{Int64}:
0
1
0
julia> using PythonCall: PyArray
julia> syndrome = PyArray(H)*error .% 2
2-element Vector{Int64}:
1
1
julia> decoding = bpd.decode(np.array(syndrome))
Python: array([0, 1, 0])