def __csd_correlation(v, m):
'''compute correlation coefficient between CSD estimate and CSD
for a given source diameter'''
if m == 'delta':
icsd_input[m].update({'diam': v * pq.m})
_icsd = icsd.DeltaiCSD(**icsd_input[m])
corrcoef = pl.corrcoef(
CSD_filtered.flatten(),
pl.array(_icsd.filter_csd(_icsd.get_csd()) / pq.m).flatten())
elif m == 'step':
icsd_input[m].update({'diam': v * pq.m})
_icsd = icsd.StepiCSD(**icsd_input[m])
corrcoef = pl.corrcoef(
CSD_filtered.flatten(),
pl.array(_icsd.filter_csd(_icsd.get_csd()) / pq.m).flatten())
elif m == 'spline':
icsd_input[m].update({'diam': v * pq.m})
_icsd = icsd.SplineiCSD(**icsd_input[m])
corrcoef = pl.corrcoef(
CSD76ptF.flatten(),
pl.array(_icsd.filter_csd(_icsd.get_csd()) / pq.m).flatten())
else:
raise Exception, 'm = %s should be either [delta, step, spline]' % m
return corrcoef[0, -1]
def matricesCorrelation(M1,
M2,
corrFunc=pearsonr,
finite=False,
posit=False,
avg=True,
**kwa):
''' Return the correlation between the two matrices, defined by the mean across
correlations for each columns. If posit=True, the correlation is done for positive matrices.
'''
if M1.shape != M2.shape:
print('Functional Connectomes shape are not the same')
return 0
N = M1.shape[0]
corr = zeros(N)
if corrFunc == corrcoef:
if posit:
for c in range(N):
corr[c] = corrcoef(where(M1[c] < 0, 0, M1[c]),
where(M2[c] < 0, 0, M2[c]))[0, 1]
else:
for c in range(N):
corr[c] = corrcoef(M1[c], M2[c])[0, 1]
elif corrFunc == pearsonr:
if posit:
for c in range(N):
corr[c] = pearsonr(where(M1[c] < 0, 0, M1[c]),
where(M2[c] < 0, 0, M2[c]))[0]
else:
for c in range(N):
corr[c] = pearsonr(M1[c], M2[c])[0]
else:
if posit:
for c in range(N):
corr[c] = corrFunc(where(M1[c] < 0, 0, M1[c]),
where(M2[c] < 0, 0, M2[c]), **kwa)
else:
for c in range(N):
corr[c] = corrFunc(M1[c], M2[c], **kwa)
if finite:
corr = where(isfinite(corr), corr, 0.)
if avg:
return corr.mean()
else:
return corr
def whatyouwant(paramin):
global dir_pri1, dir_pri2, dir_pri3
dens_moy = arange(0.02, 0.981, 0.03)
for d1 in range(len(dens_moy)):
dir_sub1 = dir_pri1 + '/Norm_%.2f_DensMoy_%.2f' % (paramin,
dens_moy[d1])
nbpatterns1 = len(os.listdir(dir_sub1))
for p1 in range(nbpatterns1):
d_patty1 = dir_sub1 + '/pattern_%.3i.npy' % p1
patty1 = np_load(d_patty1)
dir_sub2 = dir_pri2 + '/Norm_0.50_DensMoy_0.02'
nbpatterns2 = len(os.listdir(dir_sub2)) / 2
for p2 in range(nbpatterns2):
d_patty2 = dir_sub2 + '/pattern_%.3i.npy' % p2
patty2 = np_load(d_patty2)
d_patty0 = dir_sub2 + '/states_%.3i.jpeg' % p2
if (corrcoef(patty1, patty2)[0,1] > 0.7) \
and (abs(norm(patty1) - norm(patty2)) / norm(patty2) < 10.05):
direction = dir_pri3 + '/C_W%.2fD%.2fP%.3i_Pica_%.3i' % (
paramin, dens_moy[d1], p1, p2)
cmd = commands.getoutput('cp ' + d_patty2 + " " +
direction + '.npy')
toimage(
array(zip(*reversed(patty2.reshape(1, len(
patty2))))).T).save(direction + '.jpeg')
print direction, abs(norm(patty1) -
norm(patty2)) / norm(patty2)
return 1
def plot_xy(cursor, query, prefix=None, color='b', marker='.', xlog=False, ylog=False, xlabel='', ylabel='', title=''):
"""
Executes the 'query' which should return two numerical columns.
"""
cursor.execute(query)
x_list = []
y_list = []
for row in cursor:
(x, y) = row
if (x != None and y != None):
x_list.append(x)
y_list.append(y)
X = pylab.array(x_list)
Y = pylab.array(y_list)
pylab.figure()
pylab.hold(True)
pylab.plot(X, Y, color=color, marker=marker, linestyle='None')
if (xlog):
pylab.xscale('log')
if (ylog):
pylab.yscale('log')
pylab.title(title + " (R^2 = %.2f)" % pylab.corrcoef(X,Y)[0,1]**2)
pylab.xlabel(xlabel)
pylab.ylabel(ylabel)
if (prefix != None):
pylab.savefig('../res/%s.pdf' % prefix, format='pdf')
pylab.hold(False)
def whatyouwant(paramin):
global dir_pri1, dir_pri2, dir_pri3
normW2 = arange(1.4, 2.601, 0.05).tolist()
dens_moy = arange(0.02, 0.981, 0.03)
for d1 in range(len(dens_moy)):
dir_sub1 = dir_pri1 + '/Norm_%.2f_DensMoy_%.2f' % (paramin,
dens_moy[d1])
nbpatterns1 = len(os.listdir(dir_sub1))
for p1 in range(nbpatterns1):
d_patty1 = dir_sub1 + '/pattern_%.3i.npy' % p1
patty1 = np_load(d_patty1)
for w2 in range(len(normW2)):
dir_sub2 = dir_pri2 + '/Norm_%.2f' % (normW2[w2])
nbpatterns2 = len(os.listdir(dir_sub2)) / 2
for p2 in range(nbpatterns2):
d_patty2 = dir_sub2 + '/pattern_%.3i.npy' % p2
patty2 = np_load(d_patty2)
d_patty0 = dir_sub2 + '/states_%.3i.jpeg' % p2
if (corrcoef(patty1, patty2)[0,1] > 0.99) \
and (abs(norm(patty1) - norm(patty2)) / norm(patty2) < 0.0005):
direction = dir_pri3 + '/C_W%.2fD%.2fP%.3i_D_W%.2fP%.3i' % (
paramin, dens_moy[d1], p1, normW2[w2], p2)
cmd = commands.getoutput('cp ' + d_patty2 + " " +
direction + '.npy')
cmd = commands.getoutput('cp ' + d_patty0 + " " +
direction + '.jpeg')
print direction, abs(norm(patty1) -
norm(patty2)) / norm(patty2)
return 1
def correl_single_window(sbwA, sbwB):
"""sbwA and sbwB are obtained as subwindowsA[:,n,:,:], with n marching through all the windows.
Indexes - 0 -> epochs
1 -> subwindows
2 -> start/stop
"""
N0 = sbwA.shape[0]
N1 = sbwA.shape[1]
spk_countA = pylab.zeros((N0,N1))
spk_countB = pylab.zeros((N0,N1))
for n in xrange(N0): #Epochs
for m in xrange(N1): #subwindows
spk_countA[n,m] = sbwA[n,m,1] - sbwA[n,m,0]
spk_countB[n,m] = sbwB[n,m,1] - sbwB[n,m,0]
spk_countA = spk_countA.flatten() - spk_countA.mean()
spk_countB = spk_countB.flatten() - spk_countB.mean()
R = pylab.corrcoef(spk_countA, spk_countB)
if R.size:
r = R[0,1]
else:
r = 0
return r
def triSupPearsonCor(mat1, mat2, finite=False, norm=False, mask=None):
if mask == None:
mask = triSup(mat1, ind=True)
corr = corrcoef(mat1.take(mask), mat2.take(mask))[0, 1]
if finite:
return where(isfinite(corr), corr, 0.)
else:
return corr
def calcCorrcoef(widget, msg):
global g_frhist, g_behavhist, g_corrcoef
print("length of history %d\n" % len(g_frhist));
a = pylab.zeros((len(g_frhist),256+4))
i = 0
for v in g_frhist:
j=0
for u in v:
a[i,j] = u
j = j+1
i = i+1
g_corrcoef = pylab.corrcoef(pylab.transpose(a))
def patternChecking(candidats,
patterns,
out='patterns',
test='equal',
areIn=False,
opt=0.9):
''' Return the candidats that are not in patterns or the extended patterns.
out - "patterns": pattens extended with the different candidats
- "candidats": the different candidats from patterns
test - "equal": equality
- "corrcoef": Pearson correlation.
'''
if not len(patterns):
try:
candidats[0][0]
return candidats
except:
return array([candidats])
else:
if out == 'patterns': nruter = list(patterns)
elif out == 'candidats': nruter = []
if test == 'equal': simil = lambda x, y: x in y
elif test == 'corrcoef':
simil = lambda x, y: corrcoef(x, y)[0, 1:].max()
elif test == 'fPearson':
simil = lambda x, y: fPearsonCorrelation(x, y).max()
if areIn:
testFunc = lambda x, y: simil(x, y) >= opt
else:
testFunc = lambda x, y: simil(x, y) <= opt
try:
candidats[0][0]
for i in range(len(candidats)):
if testFunc(candidats[i], patterns):
nruter.append(candidats[i])
except:
if testFunc(candidats, patterns):
nruter.append(candidats)
return nruter
def __get_melody_similarity(self, np_audio_1, np_audio_2):
loud_threshold_1 = max(np_audio_1) * rs.BELOW_LOUDEST_PEAK
loud_threshold_2 = max(np_audio_2) * rs.BELOW_LOUDEST_PEAK
silence = rs.SILENCE_BETWEEN_REPEATS * rs.RATE
print('__get_melody_similarity', len(np_audio_1), len(np_audio_2))
peaks_in_np_audio_1 = self.__get_peaks(np_audio_1,
loud_threshold=loud_threshold_1,
step=silence)
peaks_in_np_audio_2 = self.__get_peaks(np_audio_2,
loud_threshold=loud_threshold_2,
step=silence)
# if quantity of peaks is different -> these are different melodies
if len(peaks_in_np_audio_1) != len(peaks_in_np_audio_2):
return 0
aligned_1, aligned_2 = self.__align_melodies(np_audio_1, np_audio_2)
# if the melodies' lengths differ a lot -> these are different melodies
deviation = rs.MELODY_LENGTH_DEVIATION * rs.RATE
print(
deviation,
abs(len(aligned_1) - len(aligned_2)),
)
print(len(aligned_1), len(aligned_2), '\n')
if abs(len(aligned_1) - len(aligned_2)) > deviation:
return 0
end = self.__get_min_length_from_audio_couple(aligned_1, aligned_2) - 1
trimmed_audio_1, trimmed_audio_2 = np_audio_1[:end], np_audio_2[:end]
frequencies_1 = self.__get_frequency(trimmed_audio_1)
frequencies_2 = self.__get_frequency(trimmed_audio_2)
result = pylab.corrcoef(frequencies_1, frequencies_2)
similarity = abs(float(result[1][0]))
return round(similarity, 3)
def plot_xy(cursor,
query,
prefix=None,
color='b',
marker='.',
xlog=False,
ylog=False,
xlabel='',
ylabel='',
title=''):
"""
Executes the 'query' which should return two numerical columns.
"""
cursor.execute(query)
x_list = []
y_list = []
for row in cursor:
(x, y) = row
if (x != None and y != None):
x_list.append(x)
y_list.append(y)
X = pylab.array(x_list)
Y = pylab.array(y_list)
pylab.figure()
pylab.hold(True)
pylab.plot(X, Y, color=color, marker=marker, linestyle='None')
if (xlog):
pylab.xscale('log')
if (ylog):
pylab.yscale('log')
pylab.title(title + " (R^2 = %.2f)" % pylab.corrcoef(X, Y)[0, 1]**2)
pylab.xlabel(xlabel)
pylab.ylabel(ylabel)
if (prefix != None):
pylab.savefig('../res/%s.pdf' % prefix, format='pdf')
pylab.hold(False)
def correl_single_window(sbwA, sbwB):
"""sbwA and sbwB are obtained as subwindowsA[:,n,:,:], with n marching through all the windows.
Indexes - 0 -> epochs
1 -> subwindows
2 -> start/stop
"""
N0 = sbwA.shape[0]
N1 = sbwA.shape[1]
spk_countA = pylab.zeros((N0, N1))
spk_countB = pylab.zeros((N0, N1))
for n in xrange(N0): #Epochs
for m in xrange(N1): #subwindows
spk_countA[n, m] = sbwA[n, m, 1] - sbwA[n, m, 0]
spk_countB[n, m] = sbwB[n, m, 1] - sbwB[n, m, 0]
spk_countA = spk_countA.flatten() - spk_countA.mean()
spk_countB = spk_countB.flatten() - spk_countB.mean()
R = pylab.corrcoef(spk_countA, spk_countB)
if R.size:
r = R[0, 1]
else:
r = 0
return r
for k, key in enumerate(keys):
label = f'{int(float(key[1:]))/1000}k: {results[k][-1]:0.0f}'
pl.plot(results[k], label=label, lw=3, color=colors[k])
print(label)
# pl.legend()
pl.title('Total number of infections')
pl.xlabel('Day')
pl.ylabel('Number of infections')
sc.commaticks(axis='y')
pl.subplot(1, 3, 2)
for k, key in enumerate(keys):
label = f'{int(float(key[1:]))/1000}k: {results[k][-1]/popsizes[k]*100:0.1f}'
pl.plot(results[k] / popsizes[k] * 100, label=label, lw=3, color=colors[k])
print(label)
# pl.legend()
pl.title('Attack rate')
pl.xlabel('Day')
pl.ylabel('Attack rate (%)')
pl.subplot(1, 3, 3)
fres = [res[-1] for res in results]
pl.scatter(popsizes, fres, s=150, c=colors)
pl.title('Correlation')
pl.xlabel('Population size')
pl.ylabel('Number of infections')
sc.commaticks(axis='x')
sc.commaticks(axis='y')
print(pl.corrcoef(popsizes, fres))
def whatyouwant(paramin):
global dir_pri, dir_pri2
dens_moy = arange(0.02, 0.981, 0.03) # (33)
#dens_moy = [0.74] # (33)
for d in range(len(dens_moy)):
dir_sub = '/Norm_%.2f_DensMoy_%.2f' % (paramin, dens_moy[d])
if not os.path.exists(dir_pri + dir_sub):
print 'Pas de sous-dossier ' + dir_sub
return 0
if not os.path.exists(dir_pri2 + dir_sub):
os.mkdir(dir_pri2 + dir_sub)
if not os.path.exists(dir_pri2 + dir_sub + '/ok'):
nb_patterns = 0
err = -1
nb_de_candidats = len(os.listdir(dir_pri + dir_sub)) / 2
if nb_de_candidats < 250:
for i in range(nb_de_candidats):
candidat = dir_pri + dir_sub + '/pattern_%.3i.npy' % i
#destination = '../FilesIN/pattern_000.npy'
##destination = '../FilesIN/initialization.npy'
#cmd = commands.getoutput('cp ' + candidat + " " + destination)
if candidat.sum() == 0:
break
#print i,nb_de_candidats,paramin,dens_moy[d]
prog = "../evaCure/Main.py"
sys.argv = [prog , "-dens_moy",str(dens_moy[d]), \
'-normW',str(paramin), \
'-fileIN', str(candidat), \
'-N','66', \
'-m','1', \
'-dt','0.01', \
'-dur','100', \
'-a','0.', \
'-b','1.', \
'-theta','0.5', \
'-nper','3000', \
'-err','0.00001', \
'-GUI','0', \
'-save','0', \
'-test','2', \
'-netw','0', \
'-patt','1', \
'-init','3', \
'-who','0', \
#'-fileIN','../FilesIN/initialization.npy', \
'-dens','0.5', \
'-noise','0.', \
'-conn','5', \
'-upd','0', \
'-Dx','0.1', \
'-posit','0', \
'-ap0e0','0', \
'-mT','0', \
'-p_var','0.', \
#'-normW','2.', \
'-tau_theta','1', \
'-theta_ref','0.5', \
#'-dens_moy','0.3', \
'-priT','0', \
'-resh','1', \
'-nbcol','1', \
'-thre','0', \
'-gamma','30', \
'-derr','0.2', \
'-tau','1', \
'-intFDx','0']
execfile(prog)
#Network.closers()
#coooo = corrcoef(np_load(candidat), Network.states)[0,1]
#err = 1. - where(isfinite(coooo), coooo, 0)
#print i,nb_de_candidats,paramin,dens_moy[d] , coooo, err
err = 1 - corrcoef(np_load(candidat), Network.states)[0, 1]
if err < 1e-3:
#if Network.dpos < 1e-3:
destination2 = dir_pri2 + dir_sub + '/pattern_%.3i.npy' % nb_patterns
nb_patterns += 1
cmd = commands.getoutput('cp ' + candidat + " " +
destination2)
#print 'Norm %.2f, DensMoy %.2f, candidats %i, retenu %i, dcorr %.2e' %(paramin,dens_moy[d],nb_de_candidats,nb_patterns,Network.dpos)
cmd = commands.getoutput('touch ' + dir_pri2 + dir_sub + '/ok')
print 'Norm %.2f, DensMoy %.2f, candidats %i, retenu %i, dcorr %.2e' % (
paramin, dens_moy[d], nb_de_candidats, nb_patterns, err)
return 1
def _generate_histo_plot(self):
""" This function generates the actual histogram plot"""
fighist = pylab.figure()
nr_of_feats = len(self.indexlist)
# again, number of subplot columns is 8 at most while
# using as many rows as necessary
if nr_of_feats <= 8: nr_of_cols = nr_of_feats
else: nr_of_cols = 8
nr_of_rows = (len(self.indexlist) - 1) / 8 + 1
fighist.set_size_inches((5 * nr_of_cols, 3 * nr_of_rows))
important_features = dict()
important_feature_names = pylab.array([])
itercount = 1
for feat_index in self.indexlist:
# backgroundcolor for the feature importance text
bg_color = self.own_colormap(self.normalizer(\
self.feature_time_series[tuple(feat_index)]))
fighist.add_subplot(nr_of_rows, nr_of_cols, itercount)
itercount += 1
# plot the actual histogram
pylab.hist([
self.time_series_histo[label][:, feat_index[0], feat_index[1]]
for label in self.mean_time_series.keys()
],
bins=20,
normed=True,
histtype='step')
# write feature importance as fig.text
cal = pylab.gca().axis()
pylab.text((.23 * cal[0] + .77 * cal[1]),
(.8 * cal[3] + .2 * cal[2]),
'%.2f' %
self.feature_time_series[feat_index[0], feat_index[1]],
fontsize='xx-large',
bbox=dict(fc=bg_color, ec=bg_color, alpha=0.6, pad=14))
# Title uses feature name
pylab.title(
'Channel %s at %dms' %
(self.channel_names[feat_index[1]], float(feat_index[0]) /
self.sample_data.sampling_frequency * 1000),
fontsize='x-large')
# initialize, if no important features known yet
if important_features.values() == []:
for label in self.mean_time_series.keys():
important_features[label] = \
pylab.array(self.time_series_histo[label][:,
feat_index[0], feat_index[1]])
# stack current important feature with previous
else:
for label in self.mean_time_series.keys():
important_features[label] = pylab.vstack(
(important_features[label],
pylab.array(
self.time_series_histo[label][:, feat_index[0],
feat_index[1]])))
# memorize feature name
important_feature_names = \
pylab.append(important_feature_names, \
[('%s' % self.channel_names[feat_index[1]]).ljust(4, '_')\
+ ('%dms' % (float(feat_index[0]) / \
self.sample_data.sampling_frequency * 1000)).rjust(6, '_')])
self.corr_important_feat_names = important_feature_names
for label in important_features.keys():
self.corr_important_feats[label] = \
pylab.corrcoef(important_features[label])
# Draw the "feature development" plots of the important features
self._generate_feature_development_plots(important_features)
return fighist
import sys from pylab import show, plot, figure, subplot, mean, genfromtxt, corrcoef, legend x = genfromtxt(fname=sys.argv[1], skip_header = 1, usecols = 0 ) xi = x for col in xrange(1,7): figure(1) subplot(2,3,col ) y = genfromtxt(fname=sys.argv[1], skip_header = 1, usecols = col ) yi = y #slope calculated for b = 0 in y = ax + b a = mean(xi*yi)/mean(xi**2) corrr = corrcoef(yi, a*xi) #print 'corrr ', corrr[1][0] print 'slope '+str(col), a print corrcoef(yi, a*xi)[1][0] #def func(xi, a): # return a*xi #popt, pcov = curve_fit(func, xi, yi) #print 'popt', popt #print 'pcov', pcov line = a*xi plot(xi,line,'r-',xi,yi,'o', label = a) legend() #legend(['m ', 'r**2 '],[a, corrr]) show()
xlabel("Time (s) | Señal Procesada")
subplot(2, 1, 2)
plot(timeArray, canal1)
plot(peaksNoFiltro / fs, canalNoFiltro[peaksNoFiltro])
ylabel('Amplitude')
xlabel("Time (s) | Señal No-Procesada")
plt.show()
plot(peaksFiltrado / fs, picosProcesado_suavizado, color='k') #,linewidth=3.5)
plot(peaksFiltrado / fs, picosFiltradoEnOriginal)
plt.show()
print(corrcoef(picosFiltradoEnOriginal, picosProcesado_suavizado))
print(corrcoef(canalNoFiltro[peaksNoFiltro], picosProcesado_suavizado))
'''
print("Correlación de ubicación de puntos máximos" + str(np.corrcoef(peaksFiltrado, peaksNoFiltro)))
print("Correlación de valores de puntos máximos" + str(np.corrcoef(valuesFiltrado, valuesNoFiltro)))
posPicosFiltrado = list(peaksFiltrado)
valoresFiltrado = list(valuesFiltrado['peak_heights'])
print(type(posPicosFiltrado))
print(type(valuesFiltrado))
print(type(valoresFiltrado))
print(len(peaksFiltrado))
def whatyouwant(paramin):
global dir_pri
dir_sub = dir_pri + '/Norm_%.2f' %paramin
os.mkdir(dir_sub)
multi_dens = arange(0.02,0.981,0.03) # (33)
nb_patterns = 0
tendance =[]
for d in range(len(multi_dens)):
for i in range(100):
prog = "../evaCure/Main.py"
sys.argv = [prog , "-dens",str(multi_dens[d]), \
'-normW',str(paramin), \
'-N','66', \
'-m','1', \
'-dt','0.01', \
'-dur','100', \
'-a','0.', \
'-b','1.', \
'-theta','0.5', \
'-nper','3000', \
'-err','0.00001', \
'-GUI','0', \
'-save','0', \
'-test','1', \
'-netw','0', \
'-patt','1', \
'-init','0', \
'-who','0', \
'-fileIN','../FilesIN/initialization.npy', \
#'-dens','0.5', \
'-noise','0.', \
'-conn','5', \
'-upd','0', \
'-Dx','0.1', \
'-posit','0', \
'-ap0e0','0', \
'-mT','0', \
'-p_var','0.', \
#'-normW','2.', \
'-tau_theta','1', \
'-theta_ref','0.5', \
'-dens_moy','0.26', \
'-priT','0', \
'-resh','1', \
'-nbcol','1', \
'-thre','0', \
'-gamma','30', \
'-derr','0.2', \
'-tau','1', \
'-intFDx','0']
execfile(prog)
maybe = True
for j in range(1,nb_patterns + 1):
#if corrcoef(np_load(dir_sub+'/pattern_%.3i.npy'%(j-1)), Network.X)[0,1] > 0.90:
if corrcoef(np_load(dir_sub+'/pattern_%.3i.npy'%(j-1)), Network.states)[0,1] > 0.90:
maybe = False
tendance[j-1][d] += 1
break
if maybe:
#saveStates(dir_sub+'/', Network.X, resh=1, opt=nb_patterns)
saveStates(dir_sub+'/', Network.states, resh=1, opt=nb_patterns)
tendance.append(zeros(len(multi_dens)))
tendance[nb_patterns][d] = 1
nb_patterns += 1
print 'norm :%.2f ; dens:%.3f'%(paramin,multi_dens[d])
ofi = open(dir_sub+'/tendance.txt', 'w')
for i in range(len(multi_dens)):
ofi.write('%.2f\t' %multi_dens[i])
for j in range(len(tendance)):
ofi.write('%.3i\t' %tendance[j][i])
ofi.write('\n')
ofi.close()
return 1
import sys
from pylab import show, plot, figure, subplot, mean, genfromtxt, corrcoef, legend
x = genfromtxt(fname=sys.argv[1], skip_header=1, usecols=0)
xi = x
for col in xrange(1, 7):
figure(1)
subplot(2, 3, col)
y = genfromtxt(fname=sys.argv[1], skip_header=1, usecols=col)
yi = y
#slope calculated for b = 0 in y = ax + b
a = mean(xi * yi) / mean(xi**2)
corrr = corrcoef(yi, a * xi)
#print 'corrr ', corrr[1][0]
print 'slope ' + str(col), a
print corrcoef(yi, a * xi)[1][0]
#def func(xi, a):
# return a*xi
#popt, pcov = curve_fit(func, xi, yi)
#print 'popt', popt
#print 'pcov', pcov
line = a * xi
plot(xi, line, 'r-', xi, yi, 'o', label=a)
legend()
#legend(['m ', 'r**2 '],[a, corrr])
show()
def _generate_histo_plot(self):
""" This function generates the actual histogram plot"""
fighist = pylab.figure()
nr_of_feats = len(self.indexlist)
# again, number of subplot columns is 8 at most while
# using as many rows as necessary
if nr_of_feats <= 8: nr_of_cols = nr_of_feats
else: nr_of_cols = 8
nr_of_rows = (len(self.indexlist) - 1) / 8 + 1
fighist.set_size_inches((5 * nr_of_cols, 3 * nr_of_rows))
important_features = dict()
important_feature_names = pylab.array([])
itercount = 1
for feat_index in self.indexlist:
# backgroundcolor for the feature importance text
bg_color = self.own_colormap(self.normalizer(\
self.feature_time_series[tuple(feat_index)]))
fighist.add_subplot(nr_of_rows, nr_of_cols, itercount)
itercount += 1
# plot the actual histogram
pylab.hist([self.time_series_histo[label]
[:, feat_index[0], feat_index[1]]
for label in self.mean_time_series.keys()],
bins=20, normed=True , histtype='step')
# write feature importance as fig.text
cal = pylab.gca().axis()
pylab.text((.23 * cal[0] + .77 * cal[1]),
(.8 * cal[3] + .2 * cal[2]), '%.2f' %
self.feature_time_series[feat_index[0], feat_index[1]],
fontsize='xx-large',
bbox=dict(fc=bg_color, ec=bg_color, alpha=0.6, pad=14))
# Title uses feature name
pylab.title('Channel %s at %dms' %
(self.channel_names[feat_index[1]],
float(feat_index[0]) /
self.sample_data.sampling_frequency * 1000),
fontsize='x-large')
# initialize, if no important features known yet
if important_features.values() == []:
for label in self.mean_time_series.keys():
important_features[label] = \
pylab.array(self.time_series_histo[label][:,
feat_index[0], feat_index[1]])
# stack current important feature with previous
else:
for label in self.mean_time_series.keys():
important_features[label] = pylab.vstack(
(important_features[label],
pylab.array(self.time_series_histo[label]
[:, feat_index[0], feat_index[1]])))
# memorize feature name
important_feature_names = \
pylab.append(important_feature_names, \
[('%s' % self.channel_names[feat_index[1]]).ljust(4, '_')\
+ ('%dms' % (float(feat_index[0]) / \
self.sample_data.sampling_frequency * 1000)).rjust(6, '_')])
self.corr_important_feat_names = important_feature_names
for label in important_features.keys():
self.corr_important_feats[label] = \
pylab.corrcoef(important_features[label])
# Draw the "feature development" plots of the important features
self._generate_feature_development_plots(important_features)
return fighist
