[Exercise] Stochastic gradient descent, logistic regression

Code:

1 ImportNumPy as NP2 ImportCSV3 ImportMath as Mt4 5 defHypo (TT,XX):#hypothesis function6exp=MT.E7tmp=0.08      forIinchRange (0,4):9tmp+=tt[i]*Xx[i]TenAns=mt.pow (EXP,TMP)/(1 +Mt.pow (exp,tmp)) One     returnans A  - defGDA (Tt,iter):#Random Gradient descent -      forIinchRange (1,num+1):#[1..100] thetmp=Hypo (Tt,x[i]) -          forJinchRange (0,4):#[0..3] -Tt[j]=tt[j]+iternum (iter,i) * (y[i]-tmp) *X[i][j] -  + defLikeli (TT):#Calculate likelihood function -tmp=0.0 +      forIinchRange (1,num+1): Atm=Hypo (Tt,x[i]) attmp+= (Y[i]*mt.log (tm) + (1-y[i]) *mt.log (1-TM)) -     returntmp -  - defIternum (K,J):#The functions for calculating the iteration factor are "different iterations and different iteration factors" -TMP=0.1/(k+j+1) +0.1 -     returntmp in  -Trainfile=file ('Train.csv','RB') toTrainread=Csv.reader (trainfile) +Testfile=file ('Test.csv','RB') -Testread=Csv.reader (testfile) the  *X=np.zeros ((105,5), float)#(x, y) is a training sample $Y=np.zeros (105, float)Panax NotoginsengTx=np.zeros (5,float)#parameter -Ty=np.zeros (5, float) theDx=np.zeros ((105,5), float)#(Dx,dy) is a test sample +Dy=np.zeros (105, float) A  thenum=0 +  forLineinchTrainread: -Num=num+1 $x[num]= Line $Y[num]=x[num][4] - #x[1..100] y[1..100] - trainfile.close () the  -dnum=0;Wuyi  forLineinchTestread: theDnum+=1 -dx[dnum]= Line WuDy[dnum]=dx[dnum][4] - testfile.close () About  $  forIinchRange (1,num+1): -     Print(X[i],y[i]) - Print(" ----- ") -  forIinchRange (1,dnum+1): A     Print(Dx[i],dy[i]) +  theIter=0 -lx=99999.0 $ly=Likeli (Ty) the  while(Mt.fabs (LY-LX) >0.01): the     Print(Iter,likeli (Ty), Ty) thelx=ly the GDA (ty,iter) -Iter+=1 inly=Likeli (Ty) the  the Print("  ") About  thepart=0.5 the  forIinchRange (1,dnum+1): thetmp=Hypo (Ty,dx[i]) +     if(tmp<Part ): -ans=0 the     Else:BayiAns=1 the     Print(Dy[i],tmp,ans)

View Code

Operation Result:

1//Training Sample Process:2//(ID, likelihood function value, [parameter])3(0,-69.314718055994589, Array ([0., 0., 0., 0., 0.]))4(1, -179.19591664172614, Array ([0.47418735, 0.06857472, 0.6080708, 0.21795486, 0. ]))5(2, -121.62179146512122, Array ([0.33221938,-0.24682766, 0.97199997, 0.37909118, 0. ]))6(3, -80.005102980506564, Array ([0.22184855,-0.51616823, 1.29138288, 0.51894067, 0. ]))7(4, -52.113434710918014, Array ([0.13470815,-0.74770815, 1.57528938, 0.64244598, 0. ]))8(5, -31.590025749624512, Array ([0.04771585,-0.95553279, 1.81895741, 0.74985357, 0. ]))9(6, -17.452182808491344, Array ([-0.04598164,-1.14217819, 2.02004087, 0.84120529, 0. ]))Ten(7, -10.267976018117292, Array ([-0.1248766,-1.29754958, 2.1927993, 0.92091932, 0. ])) One(8, -6.6580675453339095, Array ([-0.18789329,-1.42677758, 2.34511257, 0.9918262, 0. ])) A(9, -4.6188250760341996, Array ([-0.24093233,-1.5375812, 2.47958712, 1.05499614, 0. ])) -( -3.3744821853240419, Array ([-0.2866801,-1.63418248, 2.59854375, 1.11126847, 0. ])) -(One, -2.5747996461385223, array ([-0.32646178,-1.71915875, 2.7042384, 1.16151128, 0. ])) the( -2.0385459675596458, Array ([-0.36122587,-1.79441754, 2.79869276, 1.20656554, 0. ])) -( -1.6650980025135245, Array ([-0.39176125,-1.86149956, 2.88364276, 1.24718908, 0. ])) -( -1.3960367006166241, Array ([-0.41873723,-1.92167844, 2.96054485, 1.28403573, 0. ])) -( -1.1961863968105035, Array ([-0.442717,-1.97600919, 3.03060539, 1.31765663, 0. ])) +( -1.0436583937366917, Array ([-0.46416987,-2.02536393, 3.09481781, 1.34851106, 0. ])) -( -0.92441126768741677, Array ([-0.48348429,-2.07046263, 3.15399946, 1.37698019, 0. ])) +( -0.82917964441403547, Array ([-0.50098067,-2.11190002, 3.20882445, 1.40338032, 0. ])) A( -0.75168763375440906, Array ([-0.51692311,-2.15016851, 3.25985117, 1.42797471, 0. ])) at( -0.68758023105857857, Array ([-0.53152965,-2.18567731, 3.3075448, 1.45098335, 0. ])) -( -0.633767252199817, Array ([-0.5449808,-2.21876813, 3.35229525, 1.47259108, 0. ])) -( -0.5880105475158135, Array ([-0.55742672,-2.2497278, 3.39443149, 1.4929541, 0. ])) -( -0.54865823718178242, Array ([-0.56899284,-2.27879848, 3.43423286, 1.51220514, 0. ])) -( -0.51446987092368923, Array ([-0.57978456,-2.30618579, 3.47193808, 1.53045774, 0. ])) -( -0.48449905631347406, Array ([-0.58989091,-2.33206535, 3.5077524, 1.54780952, 0. ])) in( -0.45801316245868812, Array ([-0.59938755,-2.35658801, 3.5418533, 1.56434493, 0. ])) -( -0.43443740873146985, Array ([-0.60833908,-2.37988406, 3.57439518, 1.58013741, 0. ])) to( -0.41331528565404385, Array ([-0.61680104,-2.40206665, 3.60551298, 1.59525118, 0. ])) +( -0.39428010255982965, Array ([-0.62482134,-2.42323451, 3.63532529, 1.60974263, 0. ])) -( -0.37703423805211778, Array ([-0.6324416,-2.44347427, 3.6639368, 1.62366156, 0. ])) the( -0.36133380365882289, Array ([-0.63969811,-2.46286221, 3.69144031, 1.6370521, 0. ])) *( -0.34697716567487236, Array ([-0.64662266,-2.48146587, 3.71791846, 1.64995354, 0. ])) $( -0.33379625351555026, Array ([-0.65324325,-2.49934526, 3.74344514, 1.662401, 0. ]))Panax Notoginseng( -0.32164990574178071, Array ([-0.65958462,-2.5165539, 3.76808662, 1.67442599, 0. ])) -( -0.31041872365028822, Array ([-0.66566872,-2.53313973, 3.79190257, 1.68605687, 0. ])) the( -0.30000105253087223, Array ([-0.67151512,-2.54914582, 3.8149469, 1.69731925, 0. ])) +    A  the//predict new Data results: +//(original category, hypothesis value, classification result) -(0.0, 0.004116287123555463, 0) $(0.0, 0.004491299234282269, 0) $(0.0, 0.001997774439620067, 0) -(0.0, 9.711727014021101e-05, 0) -(1.0, 0.9986958360885878, 1) the(1.0, 0.999907833813241, 1) -(1.0, 0.998089176390621, 1)Wuyi(1.0, 0.9999771709114254, 1) the(1.0, 0.9998452542803238, 1) -  WuProcess finished with exit code 0

View Code

Training samples and data download:

Original:http://archive.ics.uci.edu/ml/datasets/iris

CSV file: Http://files.cnblogs.com/files/pdev/stochastic_GDA.zip

[Exercise] Stochastic gradient descent, logistic regression