Encyclopedia of Sparse Graph Codes
David J.C. MacKay (http://www.inference.org.uk/mackay/)
Questions about
David MacKay's Sparse graph code resources are answered here.
This is a database of sparse graph codes.
Properties recorded include the codes' parity check matrices,
their rates and minimum distances,
their empirical performance on the binary input Gaussian noise channel,
and histograms of decoding times.
These Codes and Related Data
are made freely available under the conditions that
(1) when they are used, the authors of the codes and the data
and this archive are acknowledged; (2) the authors of the
codes and data are not liable for any inaccuracies
in the materials presented here or any failure of
the codes to work as hoped.
Please scroll down to enter the archive
Note: there are more codes that I have simulated
that have not yet made it into this archive.
If you have any requests for additions, email me. With appropriate
incentives (eg financial support for my group), I can probably help.
Key for ascii performance files: see this file.
Further information about alist format.
Note on high rate codes:
the following numbers give the highest rate codes I have made
with column weights 3 and 4.
Rate j N M
0.889 4 1998 222
0.876 4 3584 444
0.889 3 999 111
0.9098 3 998 90
0.889 3 900 100
0.8745 3 494 62
0.878 3 492 60
0.927 4 4096 300
0.936 4 4376 282
Acknowledgements :
The work of the inference group is supported by
IBM Zurich research laboratory.
Under construction
- Gallager codes
- Gallager codes with rate 1/2
- Gallager codes with rate 1/3
- Assorted Regular Gallager codes
- Gallager codes with Rate 1/4
- High rate Gallager codes (R=0.7,0.77)
- Very high rate Gallager codes (R~=0.89,0.94)
- Column Weight 3, N=495, Rate=0.87
- Column Weight 3, N=999, Rate=0.89
- Column Weight 4, N~=1998, Rate=0.89
- Column Weight 4, N~=4000, Rate~=0.9
- Column Weights less than 4, N=1998, Rate=0.89
- More high rate codes (R=.82-.93)
- Gallager-like codes and Algebraically constructed codes
- Repeat-accumulate codes (Divsalar, D., Jin, H., and McEliece, R. J., 1998)
Gallager codes
Gallager codes with rate 1/2
Column weight t=3
Column weight t=4
assorted small codes
Column weight t=5
Column weight t=6
Irregular codes with rate 1/2
Regular and Irregular codes with rate 1/2, Progressive Edge Growth (PEG) construction
Gallager codes with rate 1/3
Column weight t=2,3 or 3
Column weight t=4
Column weight 4, tiny block length
Column weight t=6
Assorted Regular Gallager codes
Gallager codes with Rate 1/4
High rate Gallager codes (R=0.7,0.77)
These codes were created for comparison with 2 DSC codes (273,191) (R~0.7) and (1057,813) (R~0.77).
(273,191) (R~0.7)
(1057,813) (R~0.77)
Very high rate Gallager codes (R~=0.89,0.94)
Column Weight 3, N=495, Rate=0.87
Note, high rate Gallager codes with column weight 3 are not expected to be good codes.
Column Weight 3, N=999, Rate=0.89
Note, high rate Gallager codes with column weight 3 are not expected to be good codes. These codes were made for use in Gallager codes over GF(16), where column weight 3 is OK.
Column Weight 4, N~=1998, Rate=0.89
Column Weight 4, N~=4000, Rate~=0.9
Column Weights less than 4, N=1998, Rate=0.89
More high rate codes (R=.82-.93)
Note, very high rate Gallager codes with column weight j=3 are not expected to be good codes. For each (N,R), a pair of codes are provided, one with j=3 and one with j=4.
Length 4095 with Rate=0.82
Length 16383 with rate=0.87
Length 32000 with rate=0.93.
Gallager-like codes and Algebraically constructed codes
Tiny regular codes
algebraic/monodec10.3 (N=15,K=5,M=10,R= 0.4)
DSC codes
Repeat-accumulate codes (Divsalar, D., Jin, H., and McEliece, R. J., 1998)
Nothing here at the moment
(under construction)
Here are the details of individual codes:
96.3.963 (N=96,K=48,M=48,R=0.5)
N: 96
M: 48
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=963 N=96 GH/spec3> GHC/96.3.963
comment: Notes in READEN
96.33.964 (N=96,K=48,M=48,R=0.5)
N: 96
M: 48
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=964 N=96 GH/spec33> GHC/96.33.964
96.3.965 (N=96,K=48,M=48,R=0.5)
N: 96
M: 48
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=965 N=96 GH/spec3> GHC/96.3.965
96.33.966 (N=96,K=48,M=48,R=0.5)
N: 96
M: 48
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=966 N=96 GH/spec33> GHC/96.33.966
96.3.967 (N=96,K=48,M=48,R=0.5)
N: 96
M: 48
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=967 N=96 GH/spec3> GHC/96.3.967
96.33.968 (N=96,K=48,M=48,R=0.5)
N: 96
M: 48
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=968 N=96 GH/spec33> GHC/96.33.968
EN/C/204.3.483 (N=204,K=102,M=102,R=0.5)
N: 204
M: 102
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=483 N=204 GH/spec3> GHC/204.3.483
204.33.484 (N=204,K=102,M=102,R=0.5)
N: 204
M: 102
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=484 N=204 GH/spec33> GHC/204.33.484
204.3.485 (N=204,K=102,M=102,R=0.5)
N: 204
M: 102
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=485 N=204 GH/spec3> GHC/204.3.485
204.33.486 (N=204,K=102,M=102,R=0.5)
N: 204
M: 102
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=486 N=204 GH/spec33> GHC/204.33.486
204.3.487 (N=204,K=102,M=102,R=0.5)
N: 204
M: 102
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=487 N=204 GH/spec3> GHC/204.3.487
204.33.488 (N=204,K=102,M=102,R=0.5)
N: 204
M: 102
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=488 N=204 GH/spec33> GHC/204.33.488
408.3.834 (N=408,K=204,M=204,R=0.5)
N: 408
M: 204
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=834 N=408 GH/spec3> GHC/408.3.834
408.33.844 (N=408,K=204,M=204,R=0.5)
N: 408
M: 204
author: D.J.C. MacKay (recreated March 2009)
R: 0.5
created: GHG.p seed=844 N=408 GH/spec33> GHC/408.33.844
408.3.854 (N=408,K=204,M=204,R=0.5)
N: 408
M: 204
author: D.J.C. MacKay (recreated March 2009)
R: 0.5
created: GHG.p seed=854 N=408 GH/spec3> GHC/408.3.854
408.33.864 (N=408,K=204,M=204,R=0.5)
N: 408
M: 204
author: D.J.C. MacKay (recreated March 2009)
R: 0.5
created: GHG.p seed=864 N=408 GH/spec33> GHC/408.33.864
408.3.874 (N=408,K=204,M=204,R=0.5)
N: 408
M: 204
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=874 N=408 GH/spec3> GHC/408.3.874
408.33.884 (N=408,K=204,M=204,R=0.5)
N: 408
M: 204
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=884 N=408 GH/spec33> GHC/408.33.884
816.3.134 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=134 N=816 GH/spec3> GHC/816.3.134
816.33.144 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=144 N=816 GH/spec33> GHC/816.33.144
816.3.154 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=154 N=816 GH/spec3> GHC/816.3.154
816.33.164 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=164 N=816 GH/spec33> GHC/816.33.164
816.3.174 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=174 N=816 GH/spec3> GHC/816.3.174
816.33.184 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=184 N=816 GH/spec33> GHC/816.33.184
252.252.3.252 (N=504,K=252,M=252,R=0.5)
N: 504
M: 252
author: D.J.C. MacKay
R: 0.5
created: code4r -cheap 1 -n 252 -t1 3 -v 0 -rho 1 -hd 3 -seed 252
504.504.3.504 (N=1008,K=504,M=504,R=0.5)
N: 1008
M: 504
author: D.J.C. MacKay
R: 0.5
t: 3
created: code4r -cheap 1 -n 504 -t1 3 -v 0 -rho 1 -hd 3 -seed 504
Margulis2640.1320.3 (N=2640,K=1320,M=1320,R=0.5)
N: 2640
M: 1320
t: 3
R: 0.5
author: Mike Postol + David MacKay
created: This is Margulis's Cayley graph construction with p=11. It has distance <= 220. It has a (12,4) near-codeword.
ram17.5 (N=4896,K=2448,M=2448,R= 0.5053)
N: 4896
M: 2448
Mtrue: 2422
Ktrue: 2474
t: 3
author: Pascal Vontobel and Joachim Rosenthal.
created: Constructed using Ramanujan graphs and ideas from Margulis. (q=17, p=5) (Allerton 2000)
ram13.11.5 (N=,K=,M=,R=)
4000.2000年3月24日3 (N=4000,K=2000,M=2000,R=0.5)
N: 4000
M: 2000
author: D.J.C. MacKay
R: 0.5
t: 3
created: code6 -n1 0 -n 4000 -t2 3 -t1 3 -v 0 -m 2000 -seed 243 -o GHC/4000.2000年3月24日3
8000.40003483 (N=8000,K=4000,M=4000,R=0.5)
N: 8000
M: 4000
author: D.J.C. MacKay
R: 0.5
t: 3
created: code6 -n1 0 -n 8000 -t2 3 -t1 3 -v 0 -m 4000 -seed 483 -o GHC/8000.40003483
10000.10000.3.631 (N=20000,K=10000,M=10000,R=0.5)
N: 20000
M: 10000
author: D.J.C. MacKay
R: 0.5
t: 3
created:
816.4.383 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=383 N=816 GH/spec4> GHC/816.4.383
816.44.434 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=434 N=816 GH/spec44> GHC/816.44.434
816.4.545 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=545 N=816 GH/spec4> GHC/816.4.545
816.44.656 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=656 N=816 GH/spec44> GHC/816.44.656
816.4.767 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=767 N=816 GH/spec4> GHC/816.4.767
816.44.878 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
t: 4
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=878 N=816 GH/spec44> GHC/816.44.878
4000.2000年4月24日4 (N=4000,K=2000,M=2000,R=0.5)
N: 4000
M: 2000
author: D.J.C. MacKay
R: 0.5
t: 4
created: code6 -n1 0 -n 4000 -t2 4 -t1 4 -v 0 -m 2000 -seed 244 -o GHC/4000.2000年4月24日4
8000.40004484 (N=8000,K=4000,M=4000,R=0.5)
N: 8000
M: 4000
author: D.J.C. MacKay
R: 0.5
t: 4
created: code6 -n1 0 -n 8000 -t2 4 -t1 4 -v 0 -m 4000 -seed 484 -o GHC/8000.40004484
96.44.443 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=548 N=96 GH/spec44> GHC/96.44.443
96.44.665 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=665 N=96 GH/spec44> GHC/96.44.665
96.44.887 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=887 N=96 GH/spec44> GHC/96.44.887
60.44.246 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=548 N=60 GH/spec44> GHC/60.44.246
816.55.134 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=434 N=816 GH/spec55> GHC/816.55.134
816.55.156 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=656 N=816 GH/spec55> GHC/816.55.156
816.55.178 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=878 N=816 GH/spec55> GHC/816.55.178
204.55.187 (N=204,K=102,M=102,R=0.5)
N: 204
M: 102
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=837 N=204 GH/spec55> GHC/204.55.187
204.55.165 (N=204,K=102,M=102,R=0.5)
N: 204
M: 102
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=521679 N=204 GH/spec55> GHC/204.55.165
204.55.153 (N=204,K=102,M=102,R=0.5)
N: 204
M: 102
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=41627 N=204 GH/spec55> GHC/204.55.153
816.66.634 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=434 N=816 GH/spec66> GHC/816.66.634
816.66.656 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=656 N=816 GH/spec66> GHC/816.66.656
816.66.678 (N=816,K=408,M=408,R=0.5)
N: 816
M: 408
author: D.J.C. MacKay
R: 0.5
created: GHG.p seed=878 N=816 GH/spec66> GHC/816.66.678
4986.93xb.329 (N=9972 ,K=4986,M=4986,R= 0.5)
author: D.J.C. MacKay
N: 9972
M: 4986
t: 3,9
comment: "Super Poisson" code labelled '93x'. Simulations reported in "Comparison of Construction of Irregular Gallager Codes" in Trans. on Comm. vol 47, oct. 1999 used up to a maximum of 200 iterations.
created: GH.p file=GH93xb.p spec93xb
GH93xb.p seed=329 M=4986> ../GHC/4986.93xb.329
4986.93y.654 (N=9972 ,K=4986,M=4986,R= 0.5)
author: D.J.C. MacKay
N: 9972
M: 4986
t: 3,9
comment: "Super Poisson code" labelled '93y'. Simulations reported in "Comparison of Construction of Irregular Gallager Codes" in Trans. on Comm. vol 47, oct. 1999 used up to a maximum of 200 iterations.
created: GH.p file=GH93y.p spec93y
GH93y.p seed=654 M=4986> ../GHC/4986.93y.654
4986.l3u.306 (N=9972 ,K=4986,M=4986,R= 0.5)
author: D.J.C. MacKay
N: 9972
M: 4986
t: 3,9
comment: 4986.l3u.306 Linear-time encodeable 'l3'
created:
4986.l93.193 (N=9972 ,K=4986,M=4986,R= 0.5)
author: D.J.C. MacKay
N: 9972
M: 4986
t: 3,9
comment: 4986.l93.193 Linear-time encodeable 'l93y'. Note this file name contains the characters l(L) and 1(one).
created:
498693128 (N=9972 ,K=4986,M=4986,R= 0.5)
author: D.J.C. MacKay
N: 9972
M: 4986
t: 3,9
comment: 498693128 "Sub-Poisson code" is labelled '93a'
created: GH93b.p seed=128 M=4986> ../GHC/4986.93.128
4986.93i.939 (N=9972 ,K=4986,M=4986,R= 0.5)
author: D.J.C. MacKay
N: 9972
M: 4986
t: 3,9
comment: 4986.93i.939 "Poisson" code, is labelled '93p'
created: Using Matthew Davey's "irreg" software: irreg -i ispec/93 -c> ../GHC/4986.93i.939
49863333 (N=9972 ,K=4986,M=4986,R= 0.5)
author: D.J.C. MacKay
N: 9972
M: 4986
t: 3
comment: 49863333 Regular weight 3 code is labelled '3'
created: GH3.p seed=333 M=4986> ../GHC/4986.3.333
498633330 (N=9972 ,K=4986,M=4986,R= 0.5)
author: D.J.C. MacKay
N: 9972
M: 4986
t: 3
comment: 498633330 Regular weight 3 code is labelled '33'
created: GH33.p seed=330 M=4986> ../GHC/4986.33.330
PEGReg252x504 (N=504 ,K=252,M=252,R= 0.5)
author: Xiao-Yu Hu, IBM Zurich research labs
N: 504
M: 252
t: 3
comment: Progressive Edge Growth construction attempts to maximize girth, and empirically gives very good codes. This is the best known regular Gallager code with these parameters (M,N,t) [Best in the sense of perfromance on AWGN, Wed 11/5/05]. The code is not quite regular in row-degree.
created: MainPEG
PEGReg504x1008 (N=1008 ,K=504,M=504,R= 0.5)
author: Xiao-Yu Hu, IBM Zurich research labs
N: 1008
M: 504
t: 3
comment: Progressive Edge Growth construction attempts to maximize girth, and empirically gives very good codes. This is the best known regular Gallager code with these parameters (M,N,t) [Best in the sense of perfromance on AWGN, Wed 11/5/05]. The code is not quite regular in row-degree.
created: MainPEG
PEGirReg252x504 (N=504 ,K=252,M=252,R= 0.5)
author: Xiao-Yu Hu, IBM Zurich research labs
N: 504
M: 252
comment: Progressive Edge Growth construction attempts to maximize girth, and empirically gives very good codes. This is the best known code with these parameters (N,M). [Best in the sense of perfromance on AWGN, Wed 11/5/05]
created: MainPEG
PEGirReg504x1008 (N=1008 ,K=504,M=504,R= 0.5)
author: Xiao-Yu Hu, IBM Zurich research labs
N: 1008
M: 504
comment: Progressive Edge Growth construction attempts to maximize girth, and empirically gives very good codes. The best known code with these parameters (N,M). [Best in the sense of perfromance on AWGN, Wed 11/5/05]
created: MainPEG
PEGirUppTriang1030x2048 (N=2048 ,K=1018,M=1030,R= 0.4971)
author: Xiao-Yu Hu, IBM Zurich research labs
N: 2048
M: 1030
comment: Progressive Edge Growth construction attempts to maximize girth, and empirically gives very good codes. This code is upper triangular, so fast-encodeable.
created: MainPEG
PEGirUppTriang518x1024 (N=1024 ,K=506,M=518,R= 0.4941)
author: Xiao-Yu Hu, IBM Zurich research labs
N: 1024
M: 518
comment: Progressive Edge Growth construction attempts to maximize girth, and empirically gives very good codes. This code is upper triangular, so fast-encodeable.
created: MainPEG
1920.1280年3月30日3 (N=1920,K=640,M=1280,R= 0.3333)
author: D.J.C. MacKay
N: 1920
M: 1280
t: 3
comment: For proposed international standard for cellular telephones
created:
816.1A3.833 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=833 N=816 GH/spec1A3> GHC/816.1A3.833
816.2A3.344 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=344 N=816 GH/spec2A3> GHC/816.2A3.344
816.1A3.455 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=455 N=816 GH/spec1A3> GHC/816.1A3.455
816.2A3.566 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=566 N=816 GH/spec2A3> GHC/816.2A3.566
816.1A3.677 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=677 N=816 GH/spec1A3> GHC/816.1A3.677
816.2A3.788 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=788 N=816 GH/spec2A3> GHC/816.2A3.788
408.1A3.832 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=832 N=408 GH/spec1A3> GHC/408.1A3.832
408.2A3.342 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=342 N=408 GH/spec2A3> GHC/408.2A3.342
408.1A3.452 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=452 N=408 GH/spec1A3> GHC/408.1A3.452
408.2A3.564 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=564 N=408 GH/spec2A3> GHC/408.2A3.564
408.1A3.675 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=675 N=408 GH/spec1A3> GHC/408.1A3.675
408.2A3.786 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=786 N=408 GH/spec2A3> GHC/408.2A3.786
408.1A4.482 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=482 N=408 GH/spec1A4> GHC/408.1A4.482
408.1A4.483 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=483 N=408 GH/spec1A4> GHC/408.1A4.483
408.1A4.485 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=485 N=408 GH/spec1A4> GHC/408.1A4.485
816.1A4.842 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=842 N=816 GH/spec1A4> GHC/816.1A4.842
816.1A4.843 (N=816 ,K=272,M=544,R= 0.3346)
N: 816
M: 544
Mtrue: 543
Ktrue: 273
author: D.J.C. MacKay
R:
comment: Obviously, since this code has all columns of even weight, it must have not-full rank.
created: GHG.p seed=843 N=816 GH/spec1A4> GHC/816.1A4.843
816.1A4.845 (N=816,K=272,M=544,R= 0.3333)
N: 816
M: 544
author: D.J.C. MacKay
R:
created: GHG.p seed=845 N=816 GH/spec1A4> GHC/816.1A4.845
60.1A4.246 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=548 N=60 GH/spec1A4> GHC/60.1A4.246
60.1A4.245 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=545 N=60 GH/spec1A4> GHC/60.1A4.245
comment: has a weight 9 codeword (and none smaller)
54.1A4.254 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=142 N=54 GH/spec1A4> GHC/54.1A4.254
54.1A4.253 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=140 N=54 GH/spec1A4> GHC/54.1A4.253
48.1A4.153 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=1153 N=48 GH/spec1A4> GHC/48.1A4.153
comment: has distance 8. This is an example of a code with (rate,distance) beyond the Gilbert bound: R_GV(8) = 1.0 - H(8.0/48.0) = 0.34997.
Ktrue: 17
48.1A4.154 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
Ktrue: 17
created: GHG.p seed=1163 N=48 tries=1000 GH/spec1A4> GHC/48.1A4.154
comment: has distance 8 This is an example of a code with (rate,distance) beyond the Gilbert bound: R_GV(8) = 1.0 - H(8.0/48.0) = 0.34997.
48.1A4.158 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
Ktrue: 17
created: GHG.p seed=1400 N=48 tries=1000 GH/spec1A4> GHC/48.1A4.158
comment: has distance greater than 8 This is an example of a code with (rate,distance) beyond the Gilbert bound: R_GV(8) = 1.0 - H(8.0/48.0) = 0.34997.
204.1A6.138 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=6245399 N=204 GH/spec1A6> GHC/204.1A6.138
204.1A6.139 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=46245399 N=204 GH/spec1A6> GHC/204.1A6.139
204.1A6.338 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=263438 N=204 GH/spec1A6> GHC/204.1A6.338
96.1A3.838 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=838 N=96 GH/spec1A3> GHC/96.1A3.838
96.2A3.343 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=343 N=96 GH/spec2A3> GHC/96.2A3.343
96.1A3.454 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=454 N=96 GH/spec1A3> GHC/96.1A3.454
96.2A3.565 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=565 N=96 GH/spec2A3> GHC/96.2A3.565
96.1A3.676 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=676 N=96 GH/spec1A3> GHC/96.1A3.676
96.2A3.787 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: GHG.p seed=787 N=96 GH/spec2A3> GHC/96.2A3.787
120.64.3.109 (N=120,K=56,M=64,R= 0.4667)
N: 120
M: 64
t: 3
author: D J C MacKay
created: code5 -n 120 -t1 3 -v 0 -m 64 -hd 3 -seed 1912 -o GHC/120.64.3.109
comment: This code has no 4-cycles in its graph.
120.64.3.110 (N=120,K=56,M=64,R= 0.4667)
N: 120
M: 64
t: 3
author: D J C MacKay
created: code5 -n 120 -t1 3 -v 0 -m 64 -hd 3 -seed 1091 -o GHC/120.64.3.110
comment: This code has no 4-cycles in its graph.
120.64.3.111 (N=120,K=56,M=64,R= 0.4667)
N: 120
M: 64
t: 3
author: D J C MacKay
created: code5 -n 120 -t1 3 -v 0 -m 64 -hd 3 -seed 1093 -o GHC/120.64.3.111
comment: This code has no 4-cycles in its graph.
271.127.3.112 (N=271,K=144,M=127,R= 0.5314)
N: 271
M: 127
t: 3
author: D J C MacKay
created: code5 -n 271 -t1 3 -v 0 -m 127 -hd 3 -seed 1094 -o GHC/271.127.3.112
comment: This code has no 4-cycles in its graph.
271.127.3.113 (N=271,K=144,M=127,R= 0.5314)
N: 271
M: 127
t: 3
author: D J C MacKay
created: code5 -n 271 -t1 3 -v 0 -m 127 -hd 3 -seed 1095 -o GHC/271.127.3.113
comment: This code has no 4-cycles in its graph.
271.127.3.114 (N=271,K=144,M=127,R= 0.5314)
N: 271
M: 127
t: 3
author: D J C MacKay
created: code5 -n 271 -t1 3 -v 0 -m 127 -hd 3 -seed 10916 -o GHC/271.127.3.114
comment: This code has no 4-cycles in its graph.
10002.3.333e (N=13298,K=3296,M=10002,R=0.5)
N: 13298
M: 10002
author: D.J.C. MacKay
R: 0.5
t: 3
created:
273.82.3.353 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: beckmann.p w=3 N=273 M=82 verbose=1 docodemod=0 T=3000 s0=50
273.82.3.364 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: beckmann.p w=3 N=273 M=82 verbose=1 docodemod=0 T=3000 s0=50
273.82.3.378 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: beckmann.p w=3 N=273 M=82 verbose=1 docodemod=0 T=3000 s0=50
273.82.4.556 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: beckmann.p w=4 N=273 M=82 verbose=1 docodemod=0 T=3000 s0=50
273.82.4.2763 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: beckmann.p w=4 N=273 M=82 verbose=1 docodemod=0 T=3000 s0=50
273.82.4.3065 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: beckmann.p w=4 N=273 M=82 verbose=1 docodemod=0 T=3000 s0=50
1057.244.3.352 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: beckmann.p w=3 N=1057 M=244 verbose=1 docodemod=0 T=200 s0=50
1057.244.3.353 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: beckmann.p w=3 N=1057 M=244 verbose=1 docodemod=0 T=200 s0=50
1057.244.3.457 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: beckmann.p w=3 N=1057 M=244 verbose=1 docodemod=0 T=200 s0=50
1057.244.4.360 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: beckmann.p w=4 N=1057 M=244 verbose=1 docodemod=0 T=200 s0=50
1057.244.4.364 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: beckmann.p w=4 N=1057 M=244 verbose=1 docodemod=0 T=200 s0=50
1057.244.4.389 (N=,K=,M=,R=)
N:
M:
author: D.J.C. MacKay
R:
created: beckmann.p w=4 N=1057 M=244 verbose=1 docodemod=0 T=200 s0=50
495.62.3.2915 (N=495,K=433,M=62,R= 0.8747)
N: 495
M: 62
t: 3
author: D.J.C. MacKay
comment: has one 4-cycle
created: see code/IBM.
999.111.3.5543 (N=999,K=888,M=111,R= 0.8889)
N: 999
M: 111
t: 3
author: D.J.C. MacKay
created:
999.111.3.5555 (N=999,K=888,M=111,R= 0.8889)
N: 999
M: 111
t: 3
author: D.J.C. MacKay
created:
999.111.3.5565 (N=999,K=888,M=111,R= 0.8889)
N: 999
M: 111
t: 3
author: D.J.C. MacKay
created:
1908.212.4.1383 (N=1908,K=1696,M=212,R= 0.8889)
N: 1908
M: 212
t: 4
author: D.J.C. MacKay
comment: has no 4-cycles
created: see code/IBM. Used beckmann.p.
1998年5月3日.2665 (N=1998,K=1776,M=222,R= 0.8889)
N: 1998
M: 222
t: 4
author: D.J.C. MacKay
created:
1998年5月3日.1392 (N=1998,K=1776,M=222,R= 0.8889)
N: 1998
M: 222
t: 4
author: D.J.C. MacKay
created:
1998年5月3日.1500 (N=1998,K=1776,M=222,R= 0.8889)
N: 1998
M: 222
t: 4
author: D.J.C. MacKay
created: READGHG has details of how this file was created. The code's been deleted in the move from SF.
s2.94.594 (N=1998,K=1776,M=222,R= 0.8889)
N: 1998
M: 222
t: 4
author: D.J.C. MacKay
created:
s2.94.494 (N=1998,K=1776,M=222,R= 0.8889)
N: 1998
M: 222
t: 4
author: D.J.C. MacKay
created:
4376.282.4.9598 (N=4376,K=4094,M=282,R= 0.9356)
N: 4376
M: 282
t: 4
author: D.J.C. MacKay
comment: This code was found by an intensive search using construction method 1A followed by manual pruning of two columns to remove the remaining overlaps greater than 1. It is the highest rate code I was able to make with column weight 4 and blocklength near 4376. It may be that this code is sailing too close to the wind, i.e., that its asymptotic performance for high snr is not as good as other Gallager codes. Nevertheless, in experiments so far this code competes well with RS and BCH codes.
3584年5月3日.4326 (N=3584,K=3140,M=444,R=0.876)
N: 3584
K: 3140
R: 0.876
t: 4
comment: Construction 1A was used.
author: D.J.C. MacKay
created:
s2.943.349 (N=1998,K=1776,M=222,R= 0.8889)
N: 1998
M: 222
t: 3-4
author: D.J.C. MacKay
created:
comment: This code has a few weight 3 columns and they spoil its performance
s2.94c.194 (N=1998,K=1776,M=222,R= 0.8889)
N: 1998
M: 222
t: 3-4
author: D.J.C. MacKay
created:
comment: This code has a few weight 3 columns and they spoil its performance
4095.737.3.101 (N=4095,K=3358,M=737,R= 0.82)
N: 4095
M: 737
t: 3
Mtrue: <= 737
Ktrue: 3358
author: D J C MacKay
created: code5 -n 4095 -t1 3 -v 0 -m 737 -hd 3 -seed 101 -o GHC/4095.737.3.101
comment: This code has no 4-cycles in its graph.
4095.738.4.102 (N=4095,K=3357,M=738,R= 0.82)
N: 4095
M: 738
t: 4
Mtrue: <= 737
Ktrue: 3358
author: D J C MacKay
created: code5 -n 4095 -t1 4 -v 0 -m 738 -hd 3 -seed 104 -o GHC/4095.738.4.102
comment: This code has no 4-cycles in its graph.
4161.731.4.352 (N=4095,K=3364,M=731,R= 0.8215)
N: 4095
M: 731
t: 4
comment: Alist file is gzipped
4161.731.4.356 (N=4095,K=3364,M=731,R= 0.8215)
N: 4095
M: 731
t: 4
comment: Alist file is gzipped
16383.2130年3月10日3 (N=16383,K=14253,M=2130,R= 0.87)
N: 16383
M: 2130
t: 3
Mtrue: <= 2130
Ktrue: 14253
author: D J C MacKay
created: code5 -n 16383 -t1 3 -v 0 -m 2130 -hd 3 -seed 101 -o GHC/16383.2130年3月10日3
comment: This code has no 4-cycles in its graph.
16383.2131年4月10日4 (N=16383,K=14252,M=2131,R= 0.87)
N: 16383
M: 2131
t: 4
Mtrue: <= 2130
Ktrue: 14253
author: D J C MacKay
created: code5 -n 16383 -t1 4 -v 0 -m 2131 -hd 3 -seed 104 -o GHC/16383.2131年4月10日4
comment: This code has no 4-cycles in its graph.
32000.2240年3月10日5 (N=32000,K=29760,M=2240,R= 0.93)
N: 32000
M: 2240
t: 3
Mtrue: <= 2240
Ktrue: 29760
author: D J C MacKay
created: code5 -n 32000 -t1 3 -v 0 -m 2240 -hd 3 -seed 101 -o GHC/32000.2240年3月10日5
comment: This code has no 4-cycles in its graph.
32000.2241年4月10日6 (N=32000,K=29759,M=2241,R= 0.93)
N: 32000
M: 2241
t: 4
Mtrue: <= 2240
Ktrue: 29760
author: D J C MacKay
created: code5 -n 32000 -t1 4 -v 0 -m 2241 -hd 3 -seed 108 -o GHC/32000.2241年4月10日6
comment: This code has no 4-cycles in its graph.
algebraic/monodec10.3 (N=15,K=5,M=10,R= 0.4)
N: 15
M: 10
tr: 3
Mtrue: 9
Ktrue: 6
d: 5
author: David MacKay
comment: This code is presented on page 221 of D J C MacKay (2003), and eq 13.41. I call it the pentagonful low-density parity-check code. The graph is called the Petersen graph, so maybe a good name for this code would be the (15,6) Petersen code. It has 12 words of weight 5.
DSC.273.82.A (N=273,K=0,M=273,R= 0.6996)
N: 273
M: 273
tr: 17
Mtrue: 82
Ktrue: 191
d: 18
author: Rainer Lucas and Marc Fossorier
comment: This code has no 4-cycles in its graph. It has many 6-cycles (as does any cyclic code). Its true value of M is 3.3 times smaller than its apparent value of M. Has distance 18.