From 99aa5e1b9e7d844c88af5d798a7896142245448a Mon Sep 17 00:00:00 2001 From: "Per.Andreas.Brodtkorb" Date: Thu, 29 Dec 2011 16:37:29 +0000 Subject: [PATCH] refactored parts of kreg_demo2 into kreg_demo3 and _get_data --- pywafo/src/wafo/kdetools.py | 142 ++++++++++++++++++------------------ 1 file changed, 70 insertions(+), 72 deletions(-) diff --git a/pywafo/src/wafo/kdetools.py b/pywafo/src/wafo/kdetools.py index d78d4f1..6a7aa89 100644 --- a/pywafo/src/wafo/kdetools.py +++ b/pywafo/src/wafo/kdetools.py @@ -3471,13 +3471,13 @@ def _logitinv(x): return 1.0/(np.exp(-x)+1) -def kreg_demo2(n=100, hs=None, symmetric=False, fun='hisj', plotlog=False): +def _get_data(n=100, symmetric=False, loc1=1.1, scale1=0.6, scale2=1.0): import scipy.stats as st - from sg_filter import SavitzkyGolay + #from sg_filter import SavitzkyGolay dist = st.norm - scale1 = 0.5 - loc1= 1.5 - norm1 = dist.pdf(-loc1, loc=-loc1, scale=scale1) + dist.pdf(-loc1, loc=loc1, scale=scale1) + + + norm1 = scale2*(dist.pdf(-loc1, loc=-loc1, scale=scale1) + dist.pdf(-loc1, loc=loc1, scale=scale1)) fun1 = lambda x : (dist.pdf(x, loc=-loc1, scale=scale1) + dist.pdf(x, loc=loc1, scale=scale1))/norm1 x = np.sort(6*np.random.rand(n,1)-3, axis=0) @@ -3486,19 +3486,39 @@ def kreg_demo2(n=100, hs=None, symmetric=False, fun='hisj', plotlog=False): #y = (np.cos(x)>2*np.random.rand(n, 1)-1).ravel() x = x.ravel() - alpha=0.05 - z0 = -_invnorm(alpha/2) - if symmetric: xi = np.hstack((x.ravel(),-x.ravel())) yi = np.hstack((y, y)) i = np.argsort(xi) x = xi[i] y = yi[i] - hisj = Kernel('gauss', fun='hisj').get_smoothing(x) - hste = Kernel('gauss', fun='hste').get_smoothing(x) + return x, y, fun1 +def kreg_demo2(n=100, hs=None, symmetric=False, fun='hisj', plotlog=False): + x,y, fun1 = _get_data(n, symmetric) + kreg_demo3(x,y,fun1, hs=None, fun='hisj', plotlog=False) + +def kreg_demo3(x,y, fun1, hs=None, fun='hisj', plotlog=False): + import scipy.stats as st + + + alpha=0.25 + z0 = -_invnorm(alpha/2) + + hs1 = Kernel('gauss', fun=fun).get_smoothing(x) + n = x.size + hx = np.median(np.abs(x-np.median(x)))/0.6745*(4.0/(3*n))**0.2 + if (y==True).any(): + hs2 = Kernel('gauss', fun=fun).get_smoothing(x[y==True]) + hy = np.median(np.abs(y-np.mean(y)))/0.6745*(4.0/(3*n))**0.2 + else: + hs2 = 4*hs1 + hy = 4*hx + + #hy2 = Kernel('gauss', fun=fun).get_smoothing(y) #kernel = Kernel('gauss',fun=fun) - hopt = (hste+hisj)/2 #kernel.get_smoothing(x) + hopt = (hs1+2*hs2)/3 #kernel.get_smoothing(x) + #hopt = sqrt(hs1*hs2) + #hopt=sqrt(hx*hy); if hs is None: hs = hopt @@ -3508,7 +3528,7 @@ def kreg_demo2(n=100, hs=None, symmetric=False, fun='hisj', plotlog=False): #reverse = np.exp xmin, xmax = x.min(), x.max() - ni = 2*int((xmax-xmin)/hopt)+1 + ni = max(2*int((xmax-xmin)/hopt)+1,5) print(ni) print(xmin, xmax) @@ -3516,38 +3536,32 @@ def kreg_demo2(n=100, hs=None, symmetric=False, fun='hisj', plotlog=False): xiii = np.linspace(xmin-hopt,xmax+hopt, 4*ni+1) from wafo.interpolate import stineman_interp - fact = stineman_interp([x.size], [0, 2000], [0.25, 1.0], yp=[0.75/2000,0.75/2000]).clip(min=0.25,max=1.3) + fact = stineman_interp([x.size], [0, 2000], [0.25, 1.0], yp=[0.75/2000,0.75/2000]).clip(min=0.9,max=1.0) print("fact=%g" % (fact)) kreg = KRegression(x, y, hs=hs*fact, p=0) - fi = kreg(xi) - f = kreg(xiii,output='plotobj', title='Kernel regression n=%d, hste=%g, hisj=%g' % (n,hste,hisj), plotflag=1) + #fi = kreg(xi) + f = kreg(xiii,output='plotobj', title='KRegr %s f=%g, n=%d, hs1=%g, hs2=%g' % (fun,fact,n,hs1,hs2), plotflag=1) c = gridcount(x, xi) - c0 = gridcount(x[y==True],xi) - + if (y==True).any(): + c0 = gridcount(x[y==True],xi) + else: + c0 = np.zeros(xi.shape) yi = np.where(c==0, 0, c0/c) - #yi[yi==0] = 1.0/(c[c!=0].min()+4)# - #yi[yi==1] = 1-1.0/(c[c!=0].min()+4) # + #yi[yi==0] = 1.0/(c[c!=0].min()+4) + #yi[yi==1] = 1-1.0/(c[c!=0].min()+4) #yi[yi==0] = fi[yi==0] #yi[yi==0] = np.exp(stineman_interp(xi[yi==0], xi[yi>0],np.log(yi[yi>0]))) - #yi[yi==0] = fun1(xi[yi==0]) - yi[yi==0] = yi[yi>0].min()/sqrt(n) + #yi[yi==0] = fun1(xi[yi==0]) + try: + yi[yi==0] = yi[yi>0].min()/sqrt(n) + except: + yi[yi==0] = 1./n yi[yi==1] =1-(1-yi[yi<1].max())/sqrt(n) - logity =forward(yi) -# plt.plot(xi, np.log(yi/(1-yi)), xi,logity,'.') -# plt.show() -# return - - - -# ymin = np.log(yi[yi>0].min())-1 -# logyi = np.log(yi).clip(min=ymin) -# plt.scatter(xi,logyi) -# plt.show() -# return - #print(logyi) + logity = forward(yi) + gkreg = KRegression(xi, logity, hs=hs*fact, xmin=xmin-hopt,xmax=xmax+hopt) fg = gkreg.eval_grid(xi,output='plotobj', title='Kernel regression', plotflag=1) @@ -3565,23 +3579,18 @@ def kreg_demo2(n=100, hs=None, symmetric=False, fun='hisj', plotlog=False): fg = gkreg.eval_grid(xiii,output='plotobj', title='Kernel regression', plotflag=1) + pi = reverse(fg.data) dx = xi[1]-xi[0] ckreg = KDE(x,hs=hs) #ci = ckreg.eval_grid_fast(xi)*n*dx - ciii = ckreg.eval_grid_fast(xiii)*dx*n*(1+symmetric) - - pi = reverse(fg.data) - - sa1 = np.sqrt(1./(ciii*pi*(1-pi))) - - plo3 = reverse(fg.data-z0*sa) - pup3 = reverse(fg.data+z0*sa) - fg.data = pi + ciii = ckreg.eval_grid_fast(xiii)*dx* x.size #n*(1+symmetric) +# sa1 = np.sqrt(1./(ciii*pi*(1-pi))) +# plo3 = reverse(fg.data-z0*sa) +# pup3 = reverse(fg.data+z0*sa) + fg.data = pi pi = f.data - - #pi = f.data # ref Casella and Berger (1990) "Statistical inference" pp444 @@ -3592,12 +3601,11 @@ def kreg_demo2(n=100, hs=None, symmetric=False, fun='hisj', plotlog=False): # Jeffreys intervall a=b=0.5 #st.beta.isf(alpha/2, x+a, n-x+b) - ab = 0.05 + ab = 0.03 pi1 = pi #fun1(xiii) pup2 = np.where(pi==1, 1, st.beta.isf(alpha/2, ciii*pi1+ab, ciii*(1-pi1)+ab)) plo2 = np.where(pi==0, 0, st.beta.isf(1-alpha/2, ciii*pi1+ab, ciii*(1-pi1)+ab)) - #pup = (pi + z0*np.sqrt(pi*(1-pi)/ciii)).clip(min=0,max=1) - #plo = (pi - z0*np.sqrt(pi*(1-pi)/ciii)).clip(min=0,max=1) + # Wilson score den = 1+(z0**2./ciii); @@ -3605,31 +3613,14 @@ def kreg_demo2(n=100, hs=None, symmetric=False, fun='hisj', plotlog=False): halfwidth=(z0*sqrt((pi1*(1-pi1)/ciii)+(z0**2/(4*(ciii**2)))))/den plo = (xc-halfwidth).clip(min=0) # wilson score pup = (xc+halfwidth).clip(max=1.0) # wilson score + #pup = (pi + z0*np.sqrt(pi*(1-pi)/ciii)).clip(min=0,max=1) # dont use + #plo = (pi - z0*np.sqrt(pi*(1-pi)/ciii)).clip(min=0,max=1) - logity[logity==-40] = np.nan - #saz = (logity-forward(stineman_interp(xi, xiii, plo)))/z0 -# slogity = smoothn(logity, robust=False) #, weight=1./saz) -# slogity2 = SavitzkyGolay(n=2, degree=2).smooth(logity) -# sa1 = sqrt(evar(logity)) -# sa = (slogity-logity).std() -# #print('estd = %g %g' % (sa,sa1)) -# -# -# plo3 = reverse(slogity-z0*sa) -# pup3 = reverse(slogity+z0*sa) -# syi = reverse(slogity) -# syi2 = reverse(slogity2) - #print(fg.data) - #fg.data = np.exp(fg.data) - - #plt.figure(2) fg.plot(label='KReg grid') f.plot(label='KRegression') labtxt = '%d CI' % (int(100*(1-alpha))) - #plt.plot(xi, syi, 'k',xi, syi2,'k--', label='smoothn') plt.fill_between(xiii, pup, plo, alpha=0.20,color='r', linestyle='--', label=labtxt) plt.fill_between(xiii, pup2, plo2,alpha = 0.20, color='b', linestyle=':',label='%d CI2' % (int(100*(1-alpha)))) - #plt.plot(xiii, 0.5*np.cos(xiii)+.5, 'r', label='True model') plt.plot(xiii, fun1(xiii), 'r', label='True model') plt.scatter(xi,yi, label='data') print('maxp = %g' % (np.nanmax(f.data))) @@ -3701,16 +3692,23 @@ def test_docstrings(): if __name__ == '__main__': import wafo.fig as fig - plt.ioff() + plt.ion() #test_docstrings() #kde_demo2() #kreg_demo1(fast=True) #kde_gauss_demo() #kreg_demo2(n=120,symmetric=True,fun='hste', plotlog=True) - for i, n in enumerate([10,100,1000,2000,4000]): - plt.figure(i) - kreg_demo2(n=n,symmetric=True,fun='hste', plotlog=False) - fig.tile(range(5)) + k = 0 + for i, n in enumerate([50, 100,300,600, 4000]): + x,y, fun1 = _get_data(n, symmetric=True,loc1=1.2, scale1=0.3, scale2=1.25) + k0 = k + for j, fun in enumerate(['hste', 'hisj', 'hstt', 'hldpi']): + plt.figure(k) + k +=1 + kreg_demo3(x,y,fun1, hs=None, fun=fun, plotlog=False) + #kreg_demo2(n=n,symmetric=True,fun='hste', plotlog=False) + fig.tile(range(k0,k)) + plt.ioff() plt.show() #test_smoothn_2d()