hichamjanati / pyldpc Goto Github PK

CodingMatrix for now does not work with H matrix in CSR form

Hi! Thanks for creating this library!

I've got this error when decoding one 128x128 RGB image after encoding/transmission.
Code rate (R) = 1/3
n = 9213 (= 3 * (width_RGB * 24bits) - 3 = 3*(128*24) - 3 = 9213
k = 3072
d_v = 2
d_c = 3
seed = 42
systematic=True and sparse=True

Could anyone please help me with this error?
Thanks in advance!

Have you considered this modified method to compute Lr in decoder?
It may offer some speedup, although I haven't tested it.

def phi0(x):
    x = abs(x)
    if (x < 9.08e-5 ):
        return( 10 );
    else:
        return -log (tanh (x/2))

def G(Lq):
    X = sum (phi0(e) for e in Lq)
    s = np.prod(np.sign(Lq))
    return s * phi0(X)

Thanks for your work. It seems that the encoder uses BPSK modulation in default. Where should I change the code to use QPSK modulation?

From the https://github.com/hichamjanati/pyldpc/blob/master/pyldpc/code.py

 d_c: int, Number of bits in the same parity-check equation. d_c Must be
        greater or equal to d_v and must divide n.

and then:

    if d_c <= d_v:
        raise ValueError("""d_c must be greater than d_v.""")

So what version is correct?

python3-numpy-1.21.5-1.fc35.x86_64

When running I get:

/home/nbecker/.local/lib/python3.10/site-packages/pyldpc/utils.py:151: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.
  bits = np.array(bits)
/home/nbecker/.local/lib/python3.10/site-packages/pyldpc/utils.py:152: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.

The function to encode a given message WITHOUT adding noise seems to be missing. How can I encode a message that is meant to be transmitted over a real medium? The medium itself will add plenty of noise, I don't need any extra added.

All the example/tutorial links return 404

Hi, thank you for creating this library.

Would LDPC encoding-decoding benefit from GPU acceleration (for large batches of messages)?

Did the funcition builds a regular Parity-Check Matrix H?
But why the Matrix i get is not a Parity-Check Matrix,such as
[1 0 1 0 0 0 0 1 1 1 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 1 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
[0 1 0 0 0 1 0 0 1 0 0 1 0 0 1]
[0 0 0 0 0 0 0 0 0 1 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 1 1 0 0 1 0 0 0 1 1]
[0 0 0 0 0 0 0 0 1 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
[0 1 0 0 0 1 1 0 1 0 0 0 1 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0 0 1 0]
[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]

Hello

I am using dv = 4, dc = 5, n = 20, which gives k = 7.
I generate a message of length 7, encode them using ldpc encode, with snr=100000 [since I don't want to add noise]. I then simulate errors in communication by randomly flipping 1 bit of the encoded message.
But, I am unable to decode this message correctly!

Here's the code:
`seed = np.random.RandomState(42)

H, G = make_ldpc(n, d_v, d_c, systematic=True, sparse=True)

input_msg = np.random.randint(2, size=k)

enc_msg = encode(G, input_msg, snr, seed=seed)

dec_msg = get_message(G, decode(H, enc_msg, snr)) # works correctly

flip_ind = np.random.randint(0, len(enc_msg) - 1 )

enc_msg[flip_ind] = -1.0 * enc_msg[flip_ind] # change 1 to -1 and vice versa, to simulate bit flip due to communication errors

flipped_dec_msg = get_message(G, decode(H, enc_msg, snr)) # this is wrong
`