Bad matrix rank error

[Matio7k]

I truly appreciate your code and want to analyze it deeper. For this reason I would like to ask about the error which I am getting in mm_qc_pega.py.

I am running the code with your default settings, nothing changed and I am getting error on this line:
assert np.linalg.matrix_rank(H) == G.n_cn, "Bad matrix rank"

Do you have an idea why the rank of matrix H is not equal the size of G matrix? To be precise, the rank of H is 1 rank smaller.

Thank you in advance!

[Lcrypto]

Depend from your weight distribution, you can get linear depended in qc representation. Try to lift it again or use different weight distribution. For example for weight 4 column regular code you ll always get not full rank.
With BR, USA

[Matio7k]

Could you give me an example of correct weight distribution for vn_degree_node and cn_degree_nodes? I am trying with weight distributions from literature, but then I get different error during PEG phase.

[Lcrypto]

Begin from almost regular like 3 3 3 4 ... 4 for example from https://github.com/Lcrypto/trapping-sets-enumeration/tree/master/Article%20An%20Modified%20Cole's%20Importance%20Sampling%20Method%20For%20Low%20Error%20Floor%20QC-LDPC%20Codes%20Construction it always full rank

[Matio7k]

Thank you very much for your help!

[Lcrypto]

You are Welcome!

[Matio7k]

Please excuse my questions but I have one more.

When I want to make a QC-code with size e.g. n=20, m=10, I get the error:
RecursionError: maximum recursion depth exceeded while calling a Python object,
which comes from recursively calling strategy1() function in peg_utils.py (line 388). Have you seen this type of behaviour? Do you know how to deal with the problem of too big call stack?

This problem only appears when making QC-code with bigger sizes, e.g. for n=10 and m=5, everything works fine.

[Lcrypto]

I don't recommend to use column weight too large and circulant larger 128. If you want large sizes best way to use
Random generator of protograph from you weight distrubion

https://github.com/Lcrypto/Density-Evolution-AWGN/blob/master/rand_proto.m

And after use simulation annealing with large emd

https://github.com/Lcrypto/Simulated-annealing-lifting-QC-LDPC

If you still want peg better use matlab peg+ace implementation
With, BR Usa