*ë «¡;csËdZ)0dkZ1dkZ3dfd„ƒYZ<dfd„ƒYZCdfd„ƒYZOdfd „ƒYZ[d fd „ƒYZld fd „ƒYZ}dfd„ƒYZ ‡dfd„ƒYZ ´d„Z Ádfd„ƒYZ cdfd„ƒYZ (dfd„ƒYZÚeƒdded„Zñeƒddded„ZAd„ZNd„ZUd „Zbd!„Zqd"„Z{dkZ|eiZ}eiZ~ei Z dS(#sqThis provides code for working with support vector machines. Support vector machine (SVM) is a supervised machine learning algorithm that has been shown to work well on a variety of classification problems. XXX describe kernel_fn, C, epsilon For more information, see: http://svm.first.gmd.de/ http://svm.research.bell-labs.com/ Classes: SVM Holds data for a Support Vector Machine. SMOTrainer Trains an SVM using Sequential Minimal Optimization (SMO). KeerthiTrainer Trains an SVM using Keerthi's extensions to SMO. TransductiveTrainer Trains a transductive SVM. LinearKernel PolynomialKernel RadialBasisFunctionKernel HyperbolicTangentKernel Functions: train Train a Support Vector Machine on some training data. trans_transductive classify Use a Support Vector Machine to classify some data. Usage: The 'train' function is provided as a user-friendly interface module. Use this function if all you want is a plain-vanilla SVM classifier. However, if you're concerned about details such as the training algorithm, you will need to use some of the classes provided. Ns LinearKernelcs&3dZ45d„Z7d„ZRS(s k(x, y) = x*ycs 56dS(N((sself((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__init__5scs78t||ƒSdS(N(s_dotsx1sx2(sselfsx1sx2((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__call__7s(s__doc__s__init__s__call__(((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys LinearKernel3s  sSparseLinearKernelcs&<dZ=>d„Z@d„ZRS(s k(x, y) = x*ycs >?dS(N((sself((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__init__>scs@At||ƒSdS(N(s _sparse_dotsx1sx2(sselfsx1sx2((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__call__@s(s__doc__s__init__s__call__(((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pysSparseLinearKernel<s  sPolynomialKernelcs2CdZDEdZFdd„ZKd„ZRS(sk(x, y) = (x*y+1)**piicsJFG|djp ||ijoHtd|i‚nJ||_dS(Nis*polynomial degree must be between 1 and %d(spsselfs_max_ps ValueError(sselfsp((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__init__Fs cs+KLt||ƒd}M||iSdS(Nf1.0(s_dotsx1sx2sssselfsp(sselfsx1sx2ss((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__call__Ks(s__doc__s_max_ps__init__s__call__(((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pysPolynomialKernelCs  sSparsePolynomialKernelcs2OdZPQdZRdd„ZWd„ZRS(sk(x, y) = (x*y+1)**piicsJRS|djp ||ijoTtd|i‚nV||_dS(Nis*polynomial degree must be between 1 and %d(spsselfs_max_ps ValueError(sselfsp((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__init__Rs cs+WXt||ƒd}Y||iSdS(Nf1.0(s _sparse_dotsx1sx2sssselfsp(sselfsx1sx2ss((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__call__Ws(s__doc__s_max_ps__init__s__call__(((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pysSparsePolynomialKernelOs  sRadialBasisFunctionKernelcs)[dZ_`dd„Zcd„ZRS(sD -(x-y)**2 e** ---------- 2*sigma**2 f1.0cs#`a||_b|d|_dS(Ni(ssigmasselfssigma2(sselfssigma((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__init__`s cswcdt||ƒ}et||ƒ }fd|i}gyht i ||ƒSWnit j ojdSnXdS(Nf2.0f0.0( s _subtractsx1sx2sx1_x2s_dotsnumsselfssigma2sdensmathsexps OverflowError(sselfsx1sx2sx1_x2sdensnum((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__call__cs(s__doc__s__init__s__call__(((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pysRadialBasisFunctionKernel[s sSparseRadialBasisFunctionKernelcs)ldZpqdd„Ztd„ZRS(sD -(x-y)**2 e** ---------- 2*sigma**2 f1.0cs#qr||_s|d|_dS(Ni(ssigmasselfssigma2(sselfssigma((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__init__qs cswtut||ƒ}vt||ƒ }wd|i}xyyt i ||ƒSWnzt j o{dSnXdS(Nf2.0f0.0( s_sparse_subtractsx1sx2sx1_x2s _sparse_dotsnumsselfssigma2sdensmathsexps OverflowError(sselfsx1sx2sdensx1_x2snum((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__call__ts(s__doc__s__init__s__call__(((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pysSparseRadialBasisFunctionKernells sHyperbolicTangentKernelcs&}dZ~€d„Zƒd„ZRS(stanh(x*y*kappa - delta)cs€||_‚||_dS(N(skappasselfsdelta(sselfskappasdelta((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__init__€s cs7ƒ„t||ƒ}…ti||i|iƒSdS(N( s_dotsx1sx2sx1x2smathstanhsselfskappasdelta(sselfsx1sx2sx1x2((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__call__ƒs(s__doc__s__init__s__call__(((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pysHyperbolicTangentKernel}s  sSVMcs2‡dZ–—eeeƒd„Z«d„ZRS(sMHolds information for a non-linear Support Vector Machine. Members: xs A list of the input vectors. ys A list of the output values. Must be either 1 or -1. alphas A list of the Lagrange multipliers for each point. b The threshold value for the machine. kernel_fn Should take 2 vectors and return a distance. * xs, ys, and alphas should be a parallel list of vectors. ** There's a special read-only member variable 'w'. For linear SVM's, this is the 'w' vector so that the decision hyperplane is at wx-b=0 cs;—˜|tjo™dgt|iƒ}nš|tjo ›d}n||f\|_|_ž||_Ÿ||_ ||_¢t|ƒt|ƒjo£td‚n¤t|ƒt|ƒjo¥td‚n§xMt t|ƒƒd§r4}¨|i|dd gjo©td‚nqüWdS(Nf0.0s!xs and ys must be parallel arrayss%xs and alphas must be parallel arraysiisOutput must be either 1 or -1( salphassNoneslensselfsxssbsyss kernel_fns ValueErrorsrangesi(sselfsxssyssalphassbs kernel_fnsi((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys__init__—s      csΫ¬|djo«­dgt|idƒ}®x}tt|ƒƒd®rd}¯xXtt|iƒƒd¯r<}°|||i||i ||i|||| (s__doc__sNones LinearKernels__init__s __getattr__(((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pysSVM‡s cs ´¹ºd}»xytt|iƒƒd»r]}¼|i|djo ½q*n¾||i||i||i|i||ƒ}q*W¿||i SdS(sGclassify(svm, x) -> score Classify x based on an SVM object. f0.0iN( ssumsrangeslenssvmsxssisalphassyss kernel_fnsxsb(ssvmsxssumsi((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pysclassify´s   :s SMOTrainercsMÁdZËÌed„Zñd„Zd„Z[d„Z`d„ZRS(sSequential Minimal Optimization Trainer This is an implementation of the SVM trainer called Sequential Minimal Optimization, described by John C. Platt: http://www.research.microsoft.com/~jplatt Methods: train Train a new SVM. c s»ÌÑÓt||Ôdgt|ƒd|ƒ|_Õ||_Ö||_ ×h|_ Úd}Ûd} Ýx=Ý|djp| o%Þd}ß| oEàx;t t|iiƒƒdàr}á||i|ƒ}q¸Wnoãxht t|iiƒƒdãrI}å|i|ii||i ƒo æqýnç||i|ƒ}qýWè|tj oé|||iƒnê| o ëd} nì|djo íd} nqpWï|iSdS(sbtrain(self, training_set, results, kernel_fn, C, epsilon, update_fn=None) -> SVM f0.0iiN(sSVMs training_setsresultsslens kernel_fnsselfssvmsepsilonsepssCs_ecaches num_changeds examine_allsrangesalphassi2s_examine_examples _is_bounds update_fnsNone( sselfs training_setsresultss kernel_fnsCsepsilons update_fns num_changedsi2s examine_all((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pystrainÌs6 "         #   c sññò|ii|}ó|ii|}ô|i|ƒ}õ||} ÷| |i jo ||i jp| |i jo |djo]út } ût }üxºtt|iiƒƒdür›} ý| |jp|i|ii| |i ƒo þq¹nÿ|i| ƒ}ti||ƒ}| t jp ||jo|}| } nq¹W| } | t j o%|i| |ƒo dSnn tidt|iiƒdƒ}  xtt|iiƒƒd rq} |i|ii| |i ƒo  qÜn |i| |ƒo dSn| dt|iiƒ} qÜWtidt|iiƒdƒ} xctt|iiƒƒdrD}|i| |ƒo dSn| dt|iiƒ} q—WndSdS(Nii(sselfssvmsyssi2sy2salphassalph2s_EsE2sr2sepssCsNonesmax_i1s max_errorsrangeslensi1s _is_boundsE1smathsfabsserrors _take_stepsrandomsrandintsi( sselfsi2sE2sE1sy2serrorsis max_errorsalph2smax_i1si1sr2((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys_examine_exampleñsF A   0   % #  !%  %cs0||jo dSn|ii}|ii||ii|f\}}|ii ||ii |f\}}|i |ƒ}|i |ƒ} ||}!|ii} "|i}#|i}%||jo4&td|||ƒ}'t|||ƒ}n1)td||ƒ}*t||||ƒ}+||jo ,dSn.|ii||||ƒ} /|ii||||ƒ}0|ii||||ƒ} 1| | d|}2|djoW3|||||}4||jo 4|}n5||jo 5|}nnJ7||| || ||| }8||| ||||| }9||||}:||||} ;||||d||| d||| ||||}<| |||d| | | d||| ||| |}>|||jo ?|}n+@|||jo A|}n C|}Et#i$||ƒ||||jo FdSnH||||} J|i&| |ƒ o:K||| || ||||| }L|} n½M|i&||ƒ o:N||| |||||| | }O|} nlQ||| || ||||| }R||| |||||| | }S||d} U| |f\|ii|<|ii|<V| |i_Wh|_)YdSdS(Nif0.0f2.0f0.5i(*si1si2sselfssvmsxsspointsalphassalph1salph2syssy1sy2s_EsE1sE2sssbsCsepssmaxsLsminsHs kernel_fnsk11sk12sk22setasa2sf1sf2sL1sH1sLobjsHobjsmathsfabssa1s _is_boundsb1sb2s_ecache(sselfsi1si2sE2sE1sHobjsetasssepssH1sk22sa1sbsk11sk12sy2sy1sL1sb2sb1sLobjsa2salph1sf2sLsHsalph2spointsCsf1((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys _take_stepsn ))        %%EE   ) - - --) csd[\|ii|ƒ o8]t|i|ii|ƒ|ii||i| SVM f0.0iis8I could not find positive and negative training examplesif2.0N(!sSVMs training_setsresultsslens kernel_fnsselfssvmsepsilonsepssCs_fcachesb_lowsb_upsNonesi_upsi_lowsisalphassyss ValueErrors num_changedsrangesi2s_examine_examplesinner_loop_successsy2salph2s_FsF2si1s _take_steps update_fnsb(sselfs training_setsresultss kernel_fnsCsepsilons update_fnsy2s num_changedsisinner_loop_successsF2si2salph2si1((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pystrainnsR "    *+#      (  cs?ª«|ii|}¬|ii|}®|i|ƒ}¯|i |ƒo°nŸ±|i |ƒp |i |ƒo ||i jo²||_ ³||_ nP´|i|ƒp |i|ƒo ||ijoµ||_¶||_n¸d}¹|i |ƒp|i |ƒp |i |ƒo0»|i|jo¼d}½|i}nn¾|i |ƒp|i|ƒp |i|ƒo0À||i joÁd}Â|i }nnÃ|o ÄdSnÇ|i |ƒo>È|i|||i joÉ|i}n Ë|i }nÌ|i||ƒSdS(Nii(sselfssvmsyssi2sy2salphassalph2s_FsF2s_in_I_0s_in_I_1s_in_I_2sb_upsi_ups_in_I_3s_in_I_4sb_lowsi_lows optimalitysi1s _take_step(sselfsi2sy2salph2si1s optimalitysF2((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys_examine_exampleªs63 3  3 3   c"s}ÎÏ||jo ÐdSnÑ|ii}Ò|ii||ii|f\}}Ó|ii ||ii |f\}}Ô|i |ƒ|i |ƒf\}}Õ||}Ö|ii}×|i} Ø|i} Ú||jo4Ûtd||| ƒ}Üt| ||ƒ}n1Þtd||ƒ}ßt| | ||ƒ}à||jo ádSnã|ii||||ƒ}ä|ii||||ƒ}å|ii||||ƒ}æ||d|}ç|djoWè|||||}é||jo é|}nê||jo ê|}nnBì|||||||}!í|||||||}î||||}ï||||}ð||!||d|||d|||||||} ñ||!||d|||d|||||||}ó| || jo ô|}n+õ| || jo ö|}n ø|}út#i$||ƒ| ||| jo ûdSný||||}ÿ||f\|ii|<|ii|<h|_&t'}} }} xt,t-|iiƒƒdrþ}|i/|ƒp|i0|ƒp |i1|ƒoI|i |ƒ} |t'jp | |jo| }|}nn |i/|ƒp|i3|ƒp |i4|ƒoI |i |ƒ}  | t'jp | | jo | }  |} nnq7W|| f\|_(|_)|| f\|_*|_+dSdS(Nif0.0f2.0f0.5i(5si1si2sselfssvmsxsspointsalphassalph1salph2syssy1sy2s_FsF1sF2sssbsCsepssmaxsLsminsHs kernel_fnsk11sk12sk22setasa2sf1sf2sL1sH1sLobjsHobjsmathsfabssa1s_fcachesNonesi_lowsi_upsb_lowsb_upsrangeslensis_in_I_0s_in_I_3s_in_I_4sfs_in_I_1s_in_I_2("sselfsi1si2si_lowsHobjsisF1setasssfsi_upsb_upsepssLobjsH1sk22sa1sbsk11sk12sb_lowsy2sy1sL1sF2sa2salph1sf2sLsHsalph2spointsCsf1((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys _take_stepÎsv ))'        !!EE   ) )  3 3 csd|ii|ƒ o8t|i|ii|ƒ|ii||i|@t||Adgt|ƒd|ƒ|_B||_C||_ D||_ E||f\|_ |_ Fh|_ Id} Jd} LxCL| djp| o+Md} N| oEOx;tt|iiƒƒdOr} P| |i| ƒ} qßWnuRxntt|iiƒƒdRrO} T|i|ii| |i| ƒƒo Uq$nV| |i| ƒ} q$WW| tj oX| | |iƒnY| o Zd} n[| djo \d} nq—W^|iSdS(strain(self, training_set, results, num_test_cases, kernel_fn, C, C_negtest, C_postest, epsilon, update_fn=None) -> SVM f0.0iiN(sSVMs training_setsresultsslens kernel_fnsselfssvmsepsilonsepssnum_test_casessCs C_negtests C_postests_ecaches num_changeds examine_allsrangesalphassi2s_examine_examples _is_bounds_Cs update_fnsNone( sselfs training_setsresultssnum_test_casess kernel_fnsCs C_negtests C_postestsepsilons update_fns num_changedsi2s examine_all((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pystrain7s: "          )   c s`a|ii|}b|ii|}c|i|ƒ}d||} f| |i jo||i |ƒjp| |i jo |djolit } jt }kxÃtt|iiƒƒdkr¤} l| |jp&|i|ii| m|i | ƒƒo nq¿no|i| ƒ}pti||ƒ}q| t jp ||jor|}s| } nq¿Wt| } u| t j o%v|i| |ƒo wdSnnytidt|iiƒdƒ} zx–tt|iiƒƒdzrw}{|i|ii| |i | ƒƒo |qën}|i| |ƒo ~dSn| dt|iiƒ} qëWtidt|iiƒdƒ} ‚xctt|iiƒƒd‚rD}ƒ|i| |ƒo „dSn…| dt|iiƒ} q¬Wn†dSdS(Nii(sselfssvmsyssi2sy2salphassalph2s_EsE2sr2sepss_CsNonesmax_i1s max_errorsrangeslensi1s _is_boundsE1smathsfabsserrors _take_stepsrandomsrandintsi( sselfsi2sE2sE1sy2serrorsis max_errorsalph2smax_i1si1sr2((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys_examine_example`sH G   #   % )  !%  %csKˆ‰||jo ŠdSn‹|ii}Œ|ii||ii|f\}}|ii ||ii |f\}}Ž|i |ƒ}|i |ƒ}||} ‘|ii} ’|i} “|i|ƒ|i|ƒf\}}•||jo4–td|||ƒ}—t|||ƒ}n1™td||ƒ}št||||ƒ}›||jo œdSnž|ii||||ƒ}Ÿ|ii||||ƒ} |ii||||ƒ} ¡|| d|}¢|djoW£|||||}¤||jo ¤|}n¥||jo ¥|}nnJ§||| ||| || }¨||| | |||| }©|| ||}ª|| ||} «||||d|||d||| | |||}¬| |||d| | |d||| | || |}®||| jo ¯|}n+°||| jo ±|}n ³|}µt%i&||ƒ| ||| jo ¶dSn¸|| ||}º|i(||ƒ o:»|||||||||| }¼|} n½½|i(||ƒ o:¾||||||||| | }¿|} nlÁ|||||||||| }Â||||||||| | }Ã||d} Å||f\|ii|<|ii|<Æ| |i_Çh|_+ÉdSdS(Nif0.0f2.0f0.5i(,si1si2sselfssvmsxsspointsalphassalph1salph2syssy1sy2s_EsE1sE2sssbsepss_CsC1sC2smaxsLsminsHs kernel_fnsk11sk12sk22setasa2sf1sf2sL1sH1sLobjsHobjsmathsfabssa1s _is_boundsb1sb2s_ecache(sselfsi1si2sE2sE1sHobjsC1sC2setasssepssH1sk22sbsk11sk12sy2sy1sL1sb2sb1sLobjsa2salph1sa1sLsalph2sHspointsf2sf1((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys _take_stepˆsn ))  '     %%EE   ) - - --) csdËÌ|ii|ƒ o8Ít|i|ii|ƒ|ii||i| SVM Train a new support vector machine. training_set is a list containing a list of numbers, each which is a vector in the training set. results is a parallel list that contains either 1 or -1, which describes the class that the training vector belongs to. kernel_fn, C, and epsilon are optional parameters used to tune the support vector machine. update_fn is an optional callback function that called at the end of every round of training. It should take 2 parameters: update_fn(num_changed, svm) s update_fnN( sKeerthiTrainerstrains training_setsresultss kernel_fnsCsepsilons update_fnssvm(s training_setsresultss kernel_fnsCsepsilons update_fnssvm((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pystrainÚs'f0.5c  s¡ñúût||üd|d|d|d|ƒ}g} xKt t |ƒƒdr2}t |||ƒ} | i| |fƒqTW| iƒtgt | ƒ}xat t | ƒƒdrH}| |\} }||jo d|| SVM XXX document this s kernel_fnsCsepsilons update_fniif1e-05if1.0N(%strains training_setsresultss kernel_fnsCsepsilons update_fnssvms test_scoressrangeslenstest_setsisclassifysscoresappendssortsNones test_resultssindexsnum_possTransductiveTrainerstrainersC_negsC_possC_testschangedstotalsrandomsrandintsoffsetsmslsgamma_msgamma_lsminsmax(s training_setsresultsstest_setsnum_poss kernel_fnsCsC_testsepsilons update_fnsgamma_lsC_negsscores test_scoresslsmschangedsissvmsindexsC_possgamma_mstotalsoffsets test_resultsstrainer((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pystrain_transductiveñsd  !           "". %  cs„AFGt|ƒt|ƒjoHtd‚nId}Jx8tt|ƒƒdJr}K|||||}qSWL|SdS(sc_dot(x, y) -> x*y Return the dot product of x and y, where x and y are lists of numbers. svectors must be same lengthf0.0iN(slensxsys ValueErrorssumsrangesi(sxsyssumsi((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys_dotAs  csfNOd}PxI|iƒdPr6}Q|i|ƒoR|||||}nqWS|SdS(Nf0.0i(ssumsxskeyssksyshas_key(sxsyskssum((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys _sparse_dotNs   !cs‘UZ[t|ƒt|ƒjo\td‚n]tgt|ƒ}^x8tt|ƒƒd^r}_|||||| x-y Return x subtract y, where x and y are lists of numbers. svectors must be same lengthiN(slensxsys ValueErrorsNonesssrangesi(sxsysssi((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys _subtractUs csÛbch}dh}exg|iƒderT}f|i|ƒog|||||| 1 or -1 Return 1/-1 depending on the sign of x. iiN(sx(sx((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys_signqs (s__doc__smathsrandoms LinearKernelsSparseLinearKernelsPolynomialKernelsSparsePolynomialKernelsRadialBasisFunctionKernelsSparseRadialBasisFunctionKernelsHyperbolicTangentKernelsSVMsclassifys SMOTrainersKeerthiTrainersTransductiveTrainersNonestrainstrain_transductives_dots _sparse_dots _subtracts_sparse_subtracts_signscSVM(smathsHyperbolicTangentKernelstrain_transductivesSVMsclassifys _sparse_dots_sparse_subtractsSparsePolynomialKernelsSparseRadialBasisFunctionKernelscSVMs _subtracts_dots SMOTrainersRadialBasisFunctionKernelsPolynomialKernelstrainsTransductiveTrainersrandoms LinearKernelsSparseLinearKernelsKeerthiTrainers_sign((sF/mit/seven/lib/python2.1/site-packages/Bio/Tools/Classification/SVM.pys?)s2       - ¢Å²P