1017.12014
`
`f'TP.eblll ~ llr DIMd .LC. MllcKJlY
`
`I " "Tt lt ~l!r A • t M I \' t
`
`lli9~Dnchmoo~m~ a-- J I. I .• I
`
`31 Jon 87 • 28FIIO
`
`lhtlp;/twrll.rallhJ.oam•c•M•e 'er'RE'ADME.hlml
`I
`
`I
`
`1111
`
`I "II 1J 1111 I I
`
`liB
`
`olio o
`
`JilL
`
`IIIIAY
`
`... 7
`
`1888 1 -
`
`iliA'
`.....
`200
`
`David J.C. MacKAy's publications an.d related stufJ
`
`• Some d0<l11111eD1B C8l1 be obtained DDt ODly ft'om my UK web server but also tl:om a mSrml: ill Toronto. Please click
`approflria&ely. The most up to date vmion of this me iB always hen: io the UK.
`
`LutJDDdified: '11m May 617:41:071999
`
`• Most of these files m • .ps.gz", i.e. poa13cript files (.ps) c:OID)IIessed with gzip (.gz). If you don't have guuzip em your
`~D~&hiDe, you c:m find puzip exccutables far various syataDB bm; 1he msdos executlble is alao a .
`
`Otherttutl'
`• Soun:e colk fQt Bg,vcajM ncpm) II$ and otluz1hjnp· c rode lATeX, BibTeX, Pc:d; data w
`• Emr Cotre!ltlnr Code man;es
`• Pn!blapl PDP( tbil •n:l!tve?
`
`Other pJaca you mipt want tD p from here:
`• My home uaae.
`• My 1995 COline oa Ipfqrmatjgn Tl!coa I minor Q'orontq) I
`• My 1997 tutbook OD Ipformat$on Dcor;y. Pattern ''ffill'itiOD ud Neural Nebrorb
`• FrulumtlY Mbd Quntiopa allo.t Bayesian metllodt for NnnJ Networlui.l lllirl.Qr. l
`• I My mao'• COJiliDDDal pulllj .... *iop•paa- 1 Badfqrd Neal'• ftp ardaiye. l
`
`Papers arranged by topie
`
`• IpfongatioD, 'Theory. Cosljae Theory. Cryptualni!!
`
`o Icxtbggk gp Ipfqnpilioo thcgty Ipfim;pM pd l,cmtWI AfagriJbma
`0 Crii!DICT NS apd DpM qxlg
`o Early woxk on Ceypamtlvsis wd Codinr
`
`• Probabi.linic Dab Modellin1
`
`o PbDtbvjs
`o Tex!hogk gn Inform!!lillll theoxy Ipfimlnc,o !!lid La!!IDinr; Afigritbms
`G B&yesjp mctiJMs quickjn
`o Jkye!!lrm metbodi for n!!jl!l'liJ netwplb
`a ImpJemont!ltlgn ofJms!an metbom: Mon1e Carlo metbodi
`a Qausslrm procesm
`a RYn!!!!!icaJ Neural networ!g!
`a AJiipljcatjQ1!3 ofBayeajap neuralll!ltw9ru
`o ''Dt"IU mnd§llige
`a Rj#m MorDvmode1s
`a Protein seqpmce mpdellin& gpd DensitY Netwod!:s
`0 Modelling ofjm•as 'Pd pdq
`o D:ata MqdoJUna fw NcumKiC'KC
`
`• CompptltionaJ Neawcience
`
`1111
`
`Hughes, Exh. 1038, p. 1
`
`

`

`1on12014
`• Evolution
`
`FTP-able papers by David J.C. MacKay
`
`• Information Theory, Coding Theory, Cryptanalysis
`
`o Textbook on Information theory, Inference and Learning Algorithms
`
`Textbook: Information Theory, Probability and Neural Networks I Canada mirror
`
`o Gallager codes and Turbo codes
`
`mncN,ps.a;z. (55 pages) abstract. I us mjrror. CanwJa I
`' Good Error-Correcting Codes baaed on Very Spane Matrices '.
`To appear in IEEE-IT. Also, with Radfoprd Neal, in early massively shortened form: mnc4s.ps.~.
`
`(published in 1995 book 'Cryptography and Coding' LNCS 1025.1 ps mirror. Canada I
`
`mncEL.ps.az. abstract. I ps miuor. Canada I
`' Near Shannon Limit Performance of Low Density Parity Check Codes'.
`With Radford Neal. (Very short paper, published in Elec. Lett.)
`BPTD.ps.zip (old USA master site). BPTD.ps (New USA master site). BPTD.ps.gz (UK mirror. not nee. up to
`~- BPTD.ps.&Z (Canada mirror).
`Turbo Decoding as an IDStance of Pearl's 'Belief Propagation' Algorithm.
`R. J. McEJjece. D. J. C. MacKay and J.-F. Cheng. Submitted to IEEE Journal on Selected Areas in
`Communication, September 1996.
`
`sensit.ps.gz. abstract. I <- UK I Canada -> I sensit.ps.gz. abstract.
`' Sensitivity of Low Density Parity Check Codes to Decoding Assumptions '.
`David J.C. MacKay and Christopher P. Hesketh.
`
`Submissions to lSli.21
`
`muleisit.ps.a;z. abstract. I <- UK I Canada -> I mulejsit.ps.~z . abstract.
`' Shortened Turbo Codes'. (A one-page summmy)
`
`scc.ps.gz. abstract. I <- UK I Canada -> I scc.ps.gz . abstract.
`'Decoding Shortened Cyclic Codes by BeliefPropagation '.(with Simon T. Wilson) (One-page
`summmy)
`
`&fq.ps.&Z. abstract. I <-UK I Canada-> I a;fg.ps.&Z . abstract.
`' Low Density Parity Check Codes over GF(q) '. (One-page summary) (with M.C. Davey)
`
`tcc-al.ps,a:z. no abstract. I<- UK I Canada-> I tcc-al,ps,a;z. no abstract.
`' TreUia-eonstrained codes '. (Draft)
`Brendan J. Frey and David J.C. MacKay.
`Presented at Allerton 1997.
`
`rev.ps.az. abstract. I <- UK I Canada -> I rev.ps.az. abstract.
`' A Revolution: Belief Propagation in Graphs with Cycles '.
`Brendan J. Frey and David J.C. MacKay.
`Presented at NIPS 1997.
`
`ldpc-irreg.ps.gz. abstract. I <- UK I Canada -> l ldpc-irreg.ps.gz. abstract.
`' Comparison of Constructions of Irregular Gallager Codes '.
`David MacKay, Simon Wilson and Matthew Davey.
`In Proceedings of the 1998 Allerton Conference on Communication, Control, and Computing.
`
`hltps:/lweb.archive.orglwebl199905072149221http:/lwol.ra.phy.cam.ac.uklrnackayiREADME. hlml
`
`219
`
`Hughes, Exh. 1038, p. 2
`
`

`

`1on12014
`
`FTP-able papers by David J.C. MacKay
`To appear in IEEE Transactions on Communications (submitted 30 July 1998).
`irreg-its.ps.gz. abstract. I <-UK I Canada-> I irreg-its.ps.gz. abstract.
`' Decoding Times of Irregular Gallager Codes '.
`
`ra-its.ps.gz. abstract. I <-UK 1 Canada-> 1 ra-its.ps.gz. abstract.
`'Decoding Times of Repeat--Accumulate Codes'.
`seaaate.ps.iZ. abstract. I <- UK I Canada -> I seaaate.ps.iZ. abstract.
`' Evaluation of Gallager Codes for Short Block Length and High Rate Applications '.
`encyclgpedia.ps.iZ. abstract. I <- UK I Canada -> I encyclo.pedia.ps . .:z. abstract.
`' Encyclopedia of Sparse Graph Codes '.
`Encyclopedia of Sparse Graph Codes (hypertext archive, under construction)
`Encyclo.pedia of Code Perfonnance Curves (hypertext archive, under construction)
`turbo-ldpc.ps.IU'. abstract. I <-UK I Canada-> I turbo-ldpc.ps.gz. abstract.
`' Turbo Codes are Low Density Parity Check Codes '.
`Draft. Please don't look at this yet, unless you are Brendan Frey.
`
`o Early work on Cryptanalysis and Coding
`
`fe.ps.gz. abstract I ps mirror. Canada I
`'A Free Energy Minimization Framework for Inference Problems in Modulo 2 Arithmetic'. Also a
`short version of the same paper: fes.ps.~.l ps mirror. Canada I appeared in Electronics Letters, title 'A
`Free Energy Minimization Algorithm for Decoding and Cryptanalysis'. Or a short version. including
`pseudocode in appendix: fem.ps.gz. I ps mirror. Canada I
`
`• Probabilistic Data Modelling
`
`o PhD thesis
`
`o Textbook on Information theory, Inference and Learnine Aieorithms
`
`o Bayesian methods, quickies
`
`'The pope is (probably) not an alien' I <- UK I Canada-> I html
`With Sean Eddy
`
`o Bayesian methods for neural networks
`
`network.ps.~, abstract I ps mirror. Canada I
`'Probable Networks and Plausible Predictions- A Review of Practical Bayesian Methods for
`Supervised Neural Networks'
`
`This is the final version of a review paper based on a book chapter in a Springer publication; it appeared
`as a commissioned review article in Network (IOPP).
`qpi short.ps.gz I ps mirror. Canada I
`Bayesian Non-Linear Modelling with Neural Networks
`A review paper giving a basic introduction to neural networks and then describing Bayesian methods,
`with two case studies.
`qpj4.ps.~.(791K.43 pages) abstract-I ps mjrror. Canada I
`' Bayesian Methods for Neural Networks: Theory and Applications '.
`Course notes for Neural Networks Summer School
`
`hltps:/lweb.archive.orglwebl199905072149221http:/lwol.ra.phy.cam.ac.uklrnackayiREADME. hlml
`
`319
`
`Hughes, Exh. 1038, p. 3
`
`

`

`1on12014
`
`FTP-able papers by David J.C. MacKay
`
`hme.ps.gz. up to date links are here
`'Bayesian methods for Mixtures of Experts'
`Steve Waterhouse, David MacKay and Tony Robinson
`
`ica.ps.a:z. abstract. I <- UK I Canada -> I ica.ps.~. abstract.
`' Maximum Likelihood and Covariant Algorithms for Independent Component Analysis'. (see also
`Barak Pearlmutter's work)
`Draft paper.
`
`Demonstration of Inde,pendent Component Analysis (audio files)
`by James Miskin
`
`objective.ps.i7'. abstract. I<- UK I Canada-> I objectiye.ps.i7'. abstract.
`' Seanhing for' optimal' inputs with an empirical regression model'.
`
`o Implementation ofBayesian methods; Monte Carlo methods
`
`erice.ps.gz. abstract. I<- UK I Canada-> I erice.ps.gz. abstract.
`' Introduction to Monte Carlo methods '.
`A review paper in the proceedings of an Erice summer school, ed. M.Jordan.
`
`alpha.ps.i7'. abstract I ps mirror. Canada I
`Hyperparameters: optimize, or integrate out?
`
`This paper appeared in shortened form in the 1993 Maxent proceedings. The long version
`underwent a lengthy refereeing process with Neural Computation, where it is to be published
`(1999) under the title Comparison of Approximate Methods for Handling Hyperparameters.
`
`paris.ps.gz. abstract I ps mirror. Canacla. abstract I
`' Probabilistic Networks: New Models and New Methods '.
`A paper describing an interpolation model with input-dependent noise level, implemented using
`'BUGS'. This paper is for ICANN95 in Paris. For more information about the BUGS program, see
`ftp.mrc-bsu.cam.ac.uk.
`
`ensemble.ps.gz. abstract. I ps mirror. Canada I
`' Developments in Probabilistic Modelling with Neural Networks - Ensemble Learning '.
`
`nips.ps.gz. abstract I ps mirror. Canada I
`'Ensemble Learning and Evidence Maximization'.
`
`astro.ps.gz, astro.tar I ps mirror. Canacla I
`Inferring the distance to the Virgo cluster from Cepheid data using the BUGS program.
`(Bayesian inference using Gibbs Sampling). Currently this sketched report uses mock data.
`
`laplace.ps.~. abstract. I <- UK I -> Canada l laplace,ps.~. abstract.
`' Choice of Basis for Laplace Approximation '.
`To appear in Machine Learning.
`
`delve.ps.gz. abstract. I <-UK I Canada-> I delve.ps.gz. abstract.
`' More sensitive tests in DELVE '.
`
`o Dynamical Neural Networks
`
`dynet.ps.~. I <-UK I Canada-> I cJynet.ps.~. I Abstract (in Germany)
`' A Recurrent Neural Network for Modelling Dynamical Systems'.
`by Coryn A.L. Bailer-Jones, David J.C. MacKay, Philip J. Withers
`
`o Gaussian processes
`
`hltps:/lweb.archive.orglwebl199905072149221http:/lwol.ra.phy.cam.ac.uklrnackayiREADME. hlml
`
`419
`
`Hughes, Exh. 1038, p. 4
`
`

`

`1on12014
`
`FTP-able papers by David J.C. MacKay
`gpros.ps.gz. abstract. I <-UK I Canada-> I gpros.ps.gz . abstract.
`' Efficient implementation of Gaussian processes. '.
`With M.N. Gibbs. Draft paper. was 'Efficient methods for Interpolation with Gaussian Processes'
`
`See also Mark's Gaussian process web site
`nc.ps.~. abstract. I <-UK I Canada-> I nc.ps.~. abstract.
`'Variational Gaussian Process Classifiers'.
`With M.N. Gibbs.
`
`See also Mark's Gaussian process web site
`gp.ps.gz. abstract. I <-UK I Canada-> I gp.ps.gz. abstract.
`' Gaussian Processes- A Replacement for Supervised Neural Networks?'.
`Lecture notes for a tutorial at NIPS 1997. More about Gaussian processes.
`apB.ps.~. abstract. I <-UK I Canada-> l apB.ps,gz. abstract.
`' Introduction to Gaussian Processes '.
`This is a later version of the above lecture notes, to appear in proceedings of a NATO school.
`
`o Applications of Bayesian neural networks
`
`pred.ps.~ abstract I ps mirrgr. Canada I
`Bayesian Non-linear Modeling for the Energy Prediction Competition
`
`Papers for which postscript files are not available here, but may be available in
`Materials Science
`
`H.K.D.H. Bhadeshia, D.J.C. MacKay, and L.E. Svensson.
`Impact toughness ofC-MN steel arc welds- Bayesian neural network analysis.
`Materials Science and Technology, 11(10):1046-1051, 1995.
`
`L.Gavard, H.K. D.H. Bhadeshia, D.J.C. MacKay, and S.Suzuki.
`Bayesian neural network model for austenite formation in steels.
`Materials Science and Technology, pages 453-463, 1996.
`
`H.Fujii, D.J.C. MacKay, and H.K. D.H. Bhadeshia.
`Bayesian neural network analysis of fatigue crack growth rate in nickel base superalloys.
`ISIJ International, 36(11):1373-1382, 1996.
`
`T.Cool, H.K.D.H. Bhadeshia, and D.J.C. MacKay.
`The yield and ultimate tensile strength of steel welds.
`Materials Science and Engineering A, A223: 186-200, 1997.
`
`J.Jones, J.King, H.K.D.H. Bhadeshia, andD.J.C. MacKay.
`Modelling the strength ofNickel base superalloys.
`3rd International Parsons Turbine Conference, 1995.
`
`T.Cool, H.K.D.H. Bhadeshia, and D.J.C. MacKay.
`Modelling the mechanical properties in the HAZ of power plant steels i: Bayesian neural network
`analysis of proof strength.
`In H.Cerjak, editor, Mathematical Modelling of Weld Phenomena 3, Materials Modelling Series, pages
`403-442. Institute of Materials, London, 1997.
`
`D.J.C. MacKay.
`Bayesian non-linear modelling with neural networks.
`In H.Cerjak, editor, Mathematical Modelling of Weld Phenomena 3, Materials Modelling Series, pages
`359-389. Institute of Materials, London, 1997.
`
`hltps:/lweb.archive.orglwebl199905072149221http:/lwol.ra.phy.cam.ac.uklrnackayiREADME. hlml
`
`519
`
`Hughes, Exh. 1038, p. 5
`
`

`

`1on12014
`
`FTP-able papers by David J.C. MacKay
`A.Y. Badmos, H.K.D.H. Bhadeshia, and D.J.C. MacKay.
`Neural network models for the tensile properties of mechanically alloyed ODS iron-alloys. Materials
`Science and Engineering A. Accepted for publication, 1997.
`
`F. Brun, T. Yoshida, J.D. Robson, V. Narayan, H.K.D.H. Bhadeshia and D. J. C. MacKay.
`Theoretical Design of Creep Resistant Steels. 1997 (Submitted).
`
`o Language modelling
`
`lang4.ps.gz. abstract I ps mirror. Canada I
`'A Hierarchieal Dirichlet Language Model.' with L.Peto
`
`o Hidden Markov models
`
`jordan.ps.&Z. abstract. I <- UK I Canada-> I jordan.ps.&Z. abstract.
`' Equivalence of Linear Boltzmann Chains and Hidden Markov Models '.
`
`ensemblePaper.ps.gz. abstract. I <- UK I Canada -> I ensemblePaper.ps.gz. abstract.
`' Ensemble Learning for Hidden Markov Models '.
`
`o Protein sequence modelling and Density Networks
`
`density.ps.~, abstract I ps mirror. Canada I
`'Density Networks and their application to Protein Modelling' (this version appeared in
`Maximum Entropy Proceedings)
`density97 .ps.gz. abstract. I <- UK I Canada-> I density97 .ps.gz. abstract.
`D. MacKay and M. Gibbs (1997): ' Density Networks'. (this version in proceedings of Edinburgh
`meeting, ed. Jim Kay)
`
`o Modelling of images and radar
`
`nn im decon.ps.gz, abstract I ps mirror, Canada I
`'Neural Network Image Deconvolution', by John E. Tansley, Martin J. Oldfield and David J.C.
`MacKay
`
`radar3.ps,iZ. abstract I ps mirrgr. Canada I
`'Bayesian analysis of linear phased-array radar' by A.G. Green and D.J.C. MacKay.
`i3,ps.u (494K). abstract I ps mirror. Canada I
`A. Barnett and D. MacKay: 'Bayesian Comparison of Models for Images'.
`
`o Neuroscience Data Analysis
`
`newint.ps.gz. abstract. I <-UK I Canada-> I newint,ps.gz. abstract.
`D. MacKay and R. Takeuchi: 'Interpolation models with mnltiple hyperparameters'. (Published
`in to Statistics and Computing)
`
`shuffle.ps.&Z. abstract. I <-UK I Canada-> I sbufile,ps.~. abstract.
`' A Comment on Data Shuftling '.
`by David J. C. MacKay, Christopher deCharms and Virginia R. de Sa.
`bridge.ps.gz. abstract. I <- UK I Canada-> I bric:Jae.ps.gz. abstract.
`' Model fitting as an Aid to Capacitance Compensation and Bridge Balancing in Neuronal
`Recording '.
`by David J. C. MacKay and Virginia R de Sa.
`
`hltps:/lweb.archive.orglwebl199905072149221http:/lwol.ra.phy.cam.ac.uklrnackayiREADME. hlml
`
`619
`
`Hughes, Exh. 1038, p. 6
`
`

`

`1on12014
`• Computational Neuroscience
`
`FTP-able papers by David J.C. MacKay
`
`Analysis ofLinsker's simulations ofHebbian rules. (D. J. C. MacKay and K D. Miller)
`Network 1, 257-298.
`Here is a text-only version of this paper (postscript). Here is a tar file ofiZip-ed tiff or postscript files of the
`figures. More jnfoonation.
`
`The Role of Constraints in Hebbian Learning. (Kenneth D. Miller and David J. C. MacKay)
`More jpfonnation and pojpters to a tecbpjca! 11(llOrt. postscript (lJCSF.USA) I postscript (Cambridie.lJK).
`
`Hyperacuity and Coarse Coding - a short paper that I never got round to publishing
`postscript I dYi I ps mirror. Canada I dvi mirror. Canada I
`timini.ps.&Z. I <- UK I Canada-> I timins.ps.&Z.
`' Associative memory using action potential timing 1
`Roland Muller, David MacKay and Andreas Herz
`
`•
`
`• Evolution
`
`~:ene.ps.jLZ. abstract. I <-UK I Canada-> I ~:ene.ps.jLZ. abstract.
`' Rate of Information Acquisition by a Species subjected to Natural Selection 1
`
`•
`
`• Papers for which postscript files are not available here, but may be available in Materials Science
`
`H.K.D.H. Bhadeshia, D.J.C. MacKay, and L.E. Svensson.
`Impact toughness ofC-MN steel arc welds- Bayesian neural network analysis.
`Materials Science and Technology, 11(10):1046-1051, 1995.
`
`L.Gavard, H.K. D.H. Bhadeshia, D.J.C. MacKay, and S.Suzuki.
`Bayesian neural network model for austenite formation in steels.
`Materials Science and Technology, pages 453-463, 1996.
`
`H.Fujii, D.J.C. MacKay, and H.K. D.H. Bhadeshia.
`Bayesian neural network analysis of fatigue crack growth rate in nickel base superalloys.
`ISU International, 36(11):1373-1382, 1996.
`
`T.Cool, H.K.D.H. Bhadeshia, and D.J.C. MacKay.
`The yield and ultimate tensile strength of steel welds.
`Materials Science and Engineering A, A223: 186-200, 1997.
`
`J.Jones, J.King, H.K.D.H. Bhadeshia, and D.J.C. MacKay.
`Modelling the strength of Nickel base superalloys.
`3rd International Parsons Turbine Conference, 1995.
`
`T.Cool, H.K.D.H. Bhadeshia, and D.J.C. MacKay.
`Modelling the mechanical properties in the HAZ of power plant steels i: Bayesian neural network analysis of proof strength.
`In H.Cerjak, editor, Mathematical Modelling of Weld Phenomena 3, Materials Modelling Series, pages 403-442. Institute of
`Materials, London, 1997.
`
`D.J.C. MacKay.
`Bayesian non-linear modelling with neural networks.
`In H.Cerjak, editor, Mathematical Modelling of Weld Phenomena 3, Materials Modelling Series, pages 359-389. Institute of
`Materials, London, 1997.
`
`A.Y. Badmos, H.K.D.H. Bhadeshia, and D.J.C. MacKay.
`
`hltps:/lweb.archive.orglwebl199905072149221http:/lwol.ra.phy.cam.ac.uklrnackayiREADME. hlml
`
`719
`
`Hughes, Exh. 1038, p. 7
`
`

`

`FTP-able papers by David J.C. MacKay
`1on12014
`Neural network models for the tensile properties of mechanically alloyed ODS iron-alloys. Materials Science and
`Engineering A. Accepted for publication, 1997.
`
`F. Brun, T. Yoshida, J.D. Robson, V. Narayan, H.K.D.H. Bhadeshia and D. J. C. MacKay.
`Theoretical Design of Creep Resistant Steels. 1997 (Submitted).
`
`Not yet available
`
`Automatic Relevance Determination (ARD) by MacKay and Neal.
`This paper is not available, and will perhaps never be completed. But please see pred.ps and network.ps above, and
`buy Radford Neal's book (Springer, 1996); all the information is there.
`
`Source code- LaTeX, peri, C- and data.
`
`LaTeX stuff
`
`• My BibTeX ftle bibs.bib (containing every paper I have cited). And a BibTeX ftle containing just my own
`publications macJcay.bib.
`• bkmk.p Reads in a Netscape bookmark file (html) and writes out a better format bookmark file with a load of internal
`links. Useful if you have a huge bookmark file with many folders. exlllllPle result. peri source.
`• bidslbib.p is a Perl program that will be useful to anyone who uses both BIDS and LaTeX.
`It automatically transforms information from the download format of BIDS to BibTeX format. - peri source.
`• DELATEX.p: Perl code for turning latex into ascii: REAPME file. clelatex.tar. (You could also use dvi2tty, which
`works very nicely except it ruins all your equations.)
`• LaTeX and BibTeX szyle files.
`
`Ccode
`
`• The old bigback Bayesian neural network simulator. source cocle for bj-kS. information (not up to date). I wrote
`big back to develop and test the ideas of my phd, for research purposes only. I used it to make my winning entry in the
`prediction competition. It can implement Bayesian methods for regression nets and classifiers but is a research tool
`only.
`Anyone interested in non-linear data modelling with neural networks is encouraged to look into Gaussian processes.
`• macopt: A conjugate gradient optimizer that I wrote, which only makes use of gradient information and doesn't need
`a routine that returns the function value. More information about macopt.
`• Free Energy Code l;m:. (For decoding and cryptanalysis. Please contact me before taking this code.)
`• Software for simulation ofMN codes and GL codes (almost the best error correcting codes in the world!) w:.
`(c) David MacKay and Radford Neal.
`• A couple of codes for use with the above MN/GL software can be found here: w:. Or you can get individual ftles here
`directozy. There are three ftles for each code: A, Cn and G. A is the parity check matrix, and G is the generator matrix.
`
`Data sets
`
`- Robot arm data, Classification data and Prediction competition data
`
`Images for Chromatic abberation demo 11 1 2. 1112 1 .6. 1
`
`Problems using this archive? Some questions and answers.
`
`Note :If you prefer to get documents by ftp rather than by clicking in a web browser, then you can get to most of these files
`by anonymous ftp to wol.ra.phy.cam.ac.uk, cd pub/www/mackay.
`
`Conversely, if you want to get a document by http rather than by ftp, change the address from
`ftp://wol.ra.phy.cam.ac.uk/pub/www/mackay/blah to http://wol.ra.phy.cam.ac.uk/mackay/blah.
`hltps:/lweb.archive.orglwebl199905072149221http:/lwol.ra.phy.cam.ac.uklrnackayiREADME. hlml
`
`819
`
`Hughes, Exh. 1038, p. 8
`
`

`

`f'T'P.UII paj~e~~lly DeWit J.C. MacKay
`11t1712014
`wltDJ /logged inlr1 )'0111' web odtlnss tl.llfml, I got a 1Nnch of ntmdut:ript cNmzt:tus which tzt le4SI my Nut.! cape 1.12 WQ.S
`ISot tkdphering.
`OK- fbat meau1bal your Netscapo doesn't know about uncompreuios files ending in .ps..sz and displayins d1mn
`with sJ!ostview. The 1biDg to do in that situation is to sol«:t "SaYe.As file" from tho ''FiJo• menu. saving to
`lllllle.pt.gz. then UliCOlDPMM <annrip) it by band IIIII then view the posaaipl in your mvourile IDIUiler.
`
`• I mirror my arcbive in Calwla.llwa
`
`Hughes, Exh. 1038, p. 9
`
`