def plot_super(map_obstacles, laser_obstacles, title):
f = plt.figure()
sp = f.add_subplot(111)
sp.plot(map_obstacles[:, 0], map_obstacles[:, 1], '.')
sp.plot(laser_obstacles[:, 0], laser_obstacles[:, 1], '.r')
sp.set_title(title)
plt.draw()
def plot_super(map_obstacles, laser_obstacles, title):
f = plt.figure()
sp = f.add_subplot(111)
sp.plot(map_obstacles[:, 0], map_obstacles[:, 1], '.')
sp.plot(laser_obstacles[:, 0], laser_obstacles[:, 1], '.r')
sp.set_title(title)
plt.draw()
def show(self, num, sp):
"""
Display RadioNode position in the 2D strucure
"""
x = self.position[0, num]
y = self.position[1, num]
sp.plot(x, y, 'ob')
def PlotFFTInput(self):
"""Plot Time Domain of FFT Input Slice for Last Measurement"""
fig = self.CreateFigure("FFT Input")
sp = fig.add_subplot(111)
xaxis = range(0,self.fftn)
sp.plot(xaxis,[x.real for x in self.fftia],'.-',color='b',label='I')
sp.plot(xaxis,[x.imag for x in self.fftia],'.-',color='r',label='Q')
sp.set_ylabel("Magnitude")
sp.set_xlabel("Sample")
#sp.legend(bbox_to_anchor=(1,-0.1))
sp.legend(loc=2,bbox_to_anchor=(0,-0.1),ncol=4)
plt.show()
def PlotFD(self,dbfs=True):
"""Plot Frequency Domain for Last Measurement"""
freqspectrum = np.abs(self.fftoa)
freqspectrum = np.concatenate( [freqspectrum[self.fftn/2:self.fftn],freqspectrum[0:self.fftn/2]] )
if dbfs:
zerodb = 20*np.log10(self.fftn/2)
freqspectrum = (20*np.log10(abs(freqspectrum))) - zerodb
fig = self.CreateFigure("Frequency Domain")
sp = fig.add_subplot(111)
xaxis = np.r_[0:self.fftn] * (self.Sr/self.fftn)
xaxis = np.concatenate( [(xaxis[self.fftn/2:self.fftn] - self.Sr),xaxis[0:self.fftn/2]])
sp.plot(xaxis,freqspectrum,'.-',color='b',label='Spectrum')
sp.set_ylabel("dBFS")
sp.set_xlabel("Frequency")
sp.legend(loc=2,bbox_to_anchor=(0,-0.1),ncol=4)
plt.show()
def PlotTD(self):
"""Plot Time Domain of Jack Input and Output Arrays for Last Measurement"""
fig = self.CreateFigure("Time Domain")
sp = fig.add_subplot(111)
xaxis = range(0,len(self.iIa))
sp.plot(xaxis,self.iIa,'.-',color='b',label='iI')
## 180 phase shift as complex portion is created with -1j
sp.plot(xaxis,-1*self.iQa,'.-',color='r',label='iQ')
sp.plot(xaxis,self.oIa,'.-',color='c',label='oI')
sp.plot(xaxis,self.oQa,'.-',color='m',label='oQ')
## Identify RTFrames
maxy = self.oIa.max()
sp.plot([self.rtframes,self.rtframes],[-maxy,maxy],'k-',lw=3,label='RT Frames')
## Identify Sync Index
sp.plot([self.synci+self.sync2fft,self.synci+self.sync2fft],[-maxy,maxy],'g-',lw=3,label='FFT Start')
sp.plot([self.synci+self.sync2fft+self.fftn,self.synci+self.sync2fft+self.fftn],[-maxy,maxy],'y-',lw=3,label='FFT End')
sp.set_ylabel("Magnitude")
sp.set_xlabel("Sample")
#sp.legend(bbox_to_anchor=(1,-0.1))
sp.legend(loc=2,bbox_to_anchor=(0,-0.1),ncol=7)
plt.show()
def plot_reg_bayes(df, xy, traces_ind, traces_hier, feat='no_feat', burn_ind=2000, burn_hier=None, quad=False, clr_school=True):
""" create plot for bayesian derived regression lines, no groups """
keys = traces_ind.keys()
fig, axes1d = plt.subplots(nrows=1, ncols=len(keys), sharex=True, sharey=True, figsize=(8*len(keys),8))
fig.suptitle('Bayesian hierarchical regression of pre-test vs post-test scores')
cm_cmap = cm.get_cmap('Set2')
clrs = {}
clrs['ind'] = ['#00F5FF','#006266','#00585C']
clrs['hier'] = ['#FF7538','#661f00','#572610']
point_clrs = 'cm_cmap(grp.clr)' if clr_school else 'cm_cmap(j/len(keys))'
if len(keys) == 1:
axes1d = [axes1d]
for j, (sp, key) in enumerate(zip(axes1d,keys)):
# scatterplot datapoints and subplot count title
if feat == 'no_feat':
x = df[xy['x']]
y = df[xy['y']]
for grpkey, grp in df.groupby('schoolid'):
sp.scatter(grp[xy['x']],grp[xy['y']],s=40,color=eval(point_clrs),label='{} ({})'.format(grpkey,len(grp))
,alpha=0.7,edgecolor='#333333')
if feat != 'no_feat':
x = df.loc[df[feat] == key,xy['x']]
y = df.loc[df[feat] == key,xy['y']]
for grpkey, grp in df.loc[df[feat] == key].groupby('schoolid'):
sp.scatter(grp[xy['x']],grp[xy['y']],s=40,color=eval(point_clrs),label='{} ({})'.format(grpkey,len(grp))
,alpha=0.7,edgecolor='#333333')
sp.annotate('{} ({} samples)'.format(key,len(x))
,xy=(0.5,1),xycoords='axes fraction',size=14,ha='center'
,xytext=(0,6),textcoords='offset points')
if clr_school:
sp.legend(scatterpoints=1, loc=8, ncol=1, bbox_to_anchor=(1.0, 0.35), fancybox=True, shadow=True)
# setup xlims and plot 1:1 line # BODGED the xlims
xfit = np.linspace(x.min(), x.max(), 10)
sp.plot(np.array([65,135]),np.array([65,135]),linestyle='dotted',linewidth=0.5,color='#666666')
# plot actual data mean
sp.scatter(x.mean(),y.mean(),marker='+',s=500,color='#551A8B')
# plot regression: individual
alpha = traces_ind[key]['alpha'][burn_ind:]
beta = traces_ind[key]['beta'][burn_ind:]
yfit = alpha[:, None] + beta[:, None] * xfit # <- yfit for all samples at x in xfit ind
yfit_at_xmean = alpha[:, None] + beta[:, None] * x.mean()
note = '{}\nslope: {:.2f}\nincpt: {:.2f}\nmeanx: {:.2f}\nin@mx: {:.2f}'.format('individual'
,beta.mean(), alpha.mean(), x.mean(), yfit_at_xmean.mean()-x.mean())
if quad:
gamma = traces_ind[key]['gamma'][burn_ind:]
yfit = alpha[:, None] + beta[:, None] * xfit + gamma[:, None] * xfit ** 2
yfit_at_xmean = alpha[:, None] + beta[:, None] * x.mean() + gamma[:, None] * x.mean() ** 2
note = '{}\ny={:.2f} + {:.2f}x + {:.3f}x^2\nmeanx: {:.2f}\nin@mx: {:.2f}'.format('individual'
,alpha.mean(),beta.mean(),gamma.mean(),x.mean(), yfit_at_xmean.mean()-x.mean())
mu = yfit.mean(0)
yerr_975 = np.percentile(yfit,97.5,axis=0)
yerr_025 = np.percentile(yfit,2.5,axis=0)
sp.plot(xfit, mu,linewidth=3, color=clrs['ind'][0], alpha=0.8)
sp.fill_between(xfit, yerr_025, yerr_975, color=clrs['ind'][2],alpha=0.3)
sp.annotate(note,xy=(1,0),xycoords='axes fraction',xytext=(-12,6),textcoords='offset points'
,color=clrs['ind'][1],weight='bold',size=12,ha='right',va='bottom')
# plot regression: hierarchical
if traces_hier is not None:
alpha = traces_hier['alpha'][burn_hier:,j]
beta = traces_hier['beta'][burn_hier:,j]
yfit = alpha[:, None] + beta[:, None] * xfit
yfit_at_xmean = alpha[:, None] + beta[:, None] * x.mean()
note = '{}\nslope: {:.2f}\nincpt: {:.2f}\nmeanx: {:.2f}\nin@mx: {:.2f}'.format('hierarchical'
,beta.mean(),alpha.mean(),x.mean(),yfit_at_xmean.mean()-x.mean())
# if quad:
# gamma = traces_hier['gamma'][burn_hier:,j]
# yfit = alpha[:, None] + beta[:, None] * xfit + gamma[:, None] * xfit ** 2
# note = '{}\ny={:.2f} + {:.2f}x + {:.2f}x^2'.format('individual',alpha.mean(),beta.mean(),gamma.mean())
mu = yfit.mean(0)
yerr_975 = np.percentile(yfit,97.5,axis=0)
yerr_025 = np.percentile(yfit,2.5,axis=0)
sp.plot(xfit, mu,linewidth=3, color=clrs['hier'][0], alpha=0.8)
sp.fill_between(xfit, yerr_025, yerr_975, color=clrs['hier'][2], alpha=0.3)
sp.annotate(note, xy=(0,1),xycoords='axes fraction',xytext=(100,-6),textcoords='offset points'
,color=clrs['hier'][1],weight='bold',size=12,ha='right',va='top')
plt.show()
cm_cmap = cm.get_cmap('hsv')
fig, axes1d = plt.subplots(nrows=1, ncols=1, sharex=True, figsize=(8,8))
fig.subplots_adjust(wspace=0.2)
fig.suptitle('Correlation of pre-test and post-test scores')
sp = axes1d
clr = cm_cmap(0.6)
sp.scatter(x=x,y=y,color=clr,alpha=0.6,edgecolor='#999999')
# fit and plot a new linear model
regr = linear_model.LinearRegression()
regr.fit(pd.DataFrame(x),pd.DataFrame(y))
x_prime = np.linspace(x.min(),x.max(),num=10)[:,np.newaxis]
y_hat = regr.predict(x_prime)
sp.plot(x_prime,y_hat,linewidth=2,linestyle='dashed',color='green',alpha=0.8)
ss_res = regr.residues_[0]
ss_tot = np.sum((y - np.mean(y))**2)
rsq = 1 - (ss_res/ss_tot)
sp.annotate('R^2: {:.2f}\nCoef: {:.2f}\nIntr: {:.2f}'.format(rsq,regr.coef_[0][0],regr.intercept_[0])
,xy=(1,0),xycoords='axes fraction',xytext=(-12,6),textcoords='offset points'
,color='green',weight='bold',size=12,ha='right',va='bottom')
sp.set_ylabel('Post-test Score')
sp.set_xlabel('Pre-test Score')
sp.axes.grid(True,linestyle='-',color='lightgrey')
plt.subplots_adjust(top=0.95)
plt.show()
def run(self, n_simulations=1, repeat=[], plot=False):
total_sims = len(self.simulations_raw)
if type(repeat) is int:
repeat = list(range(repeat))
print('Simulations: n =', n_simulations, ', repeat =', repeat)
for n_sim in tqdm(list(range(total_sims, total_sims + n_simulations)) +
repeat,
total=n_simulations + len(repeat)):
if n_sim not in self.simulations_raw.index:
print('\n' * 2 + '*' * 70 + '\n' + '*' * 70 +
'\nSimulation #' + str(n_sim))
obs_j, x_obs, y_obs, valid_j, x_valid, y_valid, test_j, x_test, y_test = self.new_data(
plot=plot)
self.add_simulation(n_sim, obs_j, valid_j, test_j)
print('*' * 70)
else:
obs_j, valid_j, test_j = self.simulations_raw.loc[n_sim][
'obs'], self.simulations_raw.loc[n_sim][
'valid'], self.simulations_raw.loc[n_sim]['test']
x_obs, y_obs, x_test, y_test = self.data_x[obs_j], self.data_y[
obs_j], self.data_x[test_j], self.data_y[test_j]
if valid_j is not None:
x_valid, y_valid, = self.data_x[valid_j], self.data_y[
valid_j]
else:
x_valid, y_valid = None, None
print('\n' * 2 + '*' * 60 + '\n' + '*' * 60 +
'\nRepetition #' + str(n_sim) + '\n' + '*' * 60)
for sp in tqdm(self.models, total=len(self.models)):
print('\n' * 2 + '*' * 50 + '\n' + sp.name + ' #' +
str(n_sim) + '\n' + '*' * 50)
sp.observed(x_obs, y_obs)
if plot:
print('\n' + sp.name)
tictoc = time.time()
selected, start, params = self.select_model(
sp, x_valid, y_valid)
time_params, tictoc = time.time() - tictoc, time.time()
if plot:
sp.plot(params)
sp.plot_model(params)
sp.set_params(params)
sp.set_space(x_obs, y_obs, obs_j)
scores_obs = self.calc_scores(sp, params)
time_scores_obs, tictoc = time.time() - tictoc, time.time()
if valid_j is not None:
sp.set_space(x_valid, y_valid, valid_j)
scores_valid = self.calc_scores(sp, params)
else:
scores_valid = {}
time_scores_valid, tictoc = time.time() - tictoc, time.time()
if x_valid is not None:
sp.observed(np.concatenate([x_obs, x_valid]),
np.concatenate([y_obs, y_valid]))
else:
sp.observed(x_obs, y_obs)
sp.set_space(x_test, y_test, test_j)
scores_test = self.calc_scores(sp, params)
time_scores_test, tictoc = time.time() - tictoc, time.time()
if plot:
print(scores_test, time_params, time_scores_obs,
time_scores_test)
self.add_result(n_sim, sp.name, selected, start, params,
scores_obs, scores_valid, scores_test,
time_params, time_scores_obs,
time_scores_valid, time_scores_test)
