def analyze(fig, ax, trade, tradeQuantities):
m = max(tradeQuantities)
if(m == 0):
return
normalized = [x/m for x in tradeQuantities]
s = sorted(normalized)
# c = np.cumsum(s)
x = [x/(len(tradeQuantities)-1) for x in range(0,len(tradeQuantities))]
ax.plot(x,s)
kde = scipy.stats.gaussian_kde(s)
x = [x/100 for x in range(0,101)]
y = kde(x).T
ax.plot(x,y)
mn = np.median(s)
sd = np.std(s)
mnsd = mn + 2 * sd
if(mnsd < 1):
crossing = [b for b in s > mnsd].index(True) / len(s)
xb, xt = ax.get_xlim()
yb, yt = ax.get_ylim()
polygonPoints = [[crossing, yt], [crossing, yb], [xt, yb], [xt,yt]]
polygon = Polygon(polygonPoints, True)
p = PatchCollection([polygon], alpha=0.3)
ax.add_collection(p)
map = recolourMap(normalized, trade)
fig.savefig('results/trade_' + str(trade) + '.png', bbox_inches='tight')
if(mnsd < 1):
return np.argmax(normalized)
return 0
def plot_gkde(data, *args, **kwargs):
"""Plot a gaussia kernel density estimator. *args and **kwargs will be passed
directory to pyplot.plot()"""
kde = scipy.stats.gaussian_kde(data)
lower = np.mean(data) - 3*np.std(data)
upper = np.mean(data) + 3*np.std(data)
x = np.linspace(lower, upper, 100)
y = kde(x)
pp.plot(x, y, *args, **kwargs)
def KDE(a,evalRange=None,bins=1000): assert len(a) > 1, "Can't KDE a single value" if evalRange is 'padded': padding = np.max(a)-np.min(a) x = np.linspace(np.min(a)-padding,np.max(a)+padding,bins) elif evalRange is None: x = np.linspace(np.min(a),np.max(a),bins) else: x = np.linspace(evalRange[0],evalRange[1],bins) pdf = kde(a) y = pdf(x) return x,y
def __call__(self, observations):
"""Estimate the p_observations_given_state matrix for a set of observations.
If observations is a list/array of length N, returns an array of shape
(N, S), where element [t, s] is the probability of the observation at
time t assuming the system was in fact in state s.
"""
observations = numpy.asarray(observations)
if self.continuous:
state_probabilities = [kde(observations) for kde in self.state_distributions]
else:
state_probabilities = [hist[observations] for hist in self.state_distributions]
return numpy.transpose(state_probabilities)
def decompose(pic): #核密度聚类，给出极大值、极小值点、背景颜色、聚类图层 print u'图层聚类分解中...' d0 = kde(pic.reshape(-1), bw_method=0.2)(range(256)) #核密度估计 d = np.diff(d0) d1 = np.where((d[:-1]<0)*(d[1:]>0))[0] #极小值 d1 = [0]+list(d1)+[256] d2 = np.where((d[:-1]>0)*(d[1:]<0))[0] #极大值 if d1[1] < d2[0]: d2 = [0]+list(d2) if d1[len(d1)-2] > d2[len(d2)-1]: d2 = list(d2)+[255] dc = sum(map(lambda i: d2[i]*(pic >= d1[i])*(pic < d1[i+1]), range(len(d2)))) print u'分解完成. 共%s个图层'%len(d2) return dc
def build_kde(self, inputData, columnName):
if len(inputData) < 5:
print("Please input atleast 5 data points")
return
num_div = max(5, int(len(inputData)/10))
# Generating the kde function
kde_function = kde(inputData[columnName])
# Generate the plot range
plot_range = numpy.linspace(max(self.data_matrix['returns_data']['Simple_Returns']), \
min(self.data_matrix['returns_data']['Simple_Returns']), \
num = num_div)
return {'x_kde':plot_range, 'y_kde':kde_function}
def build_kde(self, inputData, columnName):
if len(inputData) < 5:
print("Please input atleast 5 data points")
return
# Get number of data points to be plotted
num_div = max(5, int(len(inputData)/10))
# Generate the plot range
plot_range = numpy.linspace(max(inputData[columnName]), \
min(inputData[columnName]), \
num = num_div)
# Generating the kde function
kde_function = kde(inputData[columnName])
# Plotting the KDE and histogram simultaneously
# ----------------------------------------------
# Initializing the plot
fig, ax1 = plt.subplots()
# Plot secondary y-axis
inputData[columnName].hist(bins = num_div)
# Create secondary y-axis
ax2 = ax1.twinx()
# Plot the graph on primary x-axis
ax1.plot(plot_range, kde_function(plot_range), 'r-', \
plot_range, kde_function(plot_range), 'ko')
# Labeling the axis
ax1.set_xlabel('Returns')
ax1.set_ylabel('KDE', color='g')
ax2.set_ylabel('Frequency', color='b')
return {'x_kde':plot_range, 'y_kde':kde_function}
def maximize(samples):
'''
Use Kernel Density Estimation to find a continuous PDF
from the discrete sampling. Maximize that distribution.
Parameters
----------
samples : list (shape = (nsamples, ndim))
Observations drawn from the distribution which is
going to be fit.
Returns
-------
maximum : list(ndim)
Maxima of the probability distributions along each axis.
'''
from scipy.optimize import minimize
from scipy.stats import gaussian_kde as kde
# Give the samples array the proper shape.
samples = np.transpose(samples)
# Estimate the continous PDF.
estimate = kde(samples)
# Take the minimum of the estimate.
def PDF(x):
return -estimate(x)
# Initial guess on maximum is made from 50th percentile of the sample.
p0 = [np.percentile(samples[i], 50) for i in range(samples.shape[0])]
# Calculate the maximum of the distribution.
maximum = minimize(PDF, p0).x
return maximum
ptsw_compY = ptsw_data[:, 2] ptsw_compZ = ptsw_data[:, 3] # Projection of the converter components # xmin = -20. xmax = +10. ymin = -16. ymax = 0. X, Y = np.mgrid[xmin:xmax:100j, ymin:ymax:100j] positions = np.vstack([X.ravel(), Y.ravel()]) pr = np.vstack([prw_compX[:4000], prw_compY[:4000]]) pdf_2d_pr = kde(pr) Z_pr = np.reshape(pdf_2d_pr(positions).T, X.shape) pts = np.vstack([ptsw_compX[:4000], ptsw_compY[:4000]]) pdf_2d_pts = kde(pts) Z_pts = np.reshape(pdf_2d_pts(positions).T, X.shape) pts_other = np.vstack([ptsw_compX[15000:21000], ptsw_compY[15000:21000]]) pdf_2d_pts_other = kde(pts_other) Z_pts_other = np.reshape(pdf_2d_pts_other(positions).T, X.shape) nlevels = 5 levels_pr = np.linspace(0.0, Z_pr.max(), nlevels + 1)[1:] levels_pts = np.linspace(0.00, Z_pts.max(), nlevels + 1)[1:] levels_pts_other = np.linspace(0.00, Z_pts_other.max(), nlevels + 1)[1:]
def plot_contours(x, y,xlabel='', ylabel='',
input_x=None, input_y=None,
input_color=['red','blue','magenta'], fontsize=22,
title='', plot_samples=False, samples_color='gray',
contour_lw=1, savefile=None, plot_contours=True,
show_legend=False,fill_contours=True,
xmin=None,xmax=None, ymin=None, ymax=None,
input_label=None):
"""
...
"""
np.atleast_1d(input_x)
np.atleast_1d(input_y)
n = 100
points = np.array([x,y])
posterior = kde(points)
if xmin is None:
xmin=x.min()
if xmax is None:
xmax=x.max()
if ymin is None:
ymin=y.min()
if ymax is None:
ymax=y.max()
step_x = ( xmax - xmin ) / n
step_y = ( ymax - ymin ) / n
grid_pars = np.mgrid[0:n,0:n]
par_x = grid_pars[0]*step_x + xmin
par_y = grid_pars[1]*step_y + ymin
grid_posterior = grid_pars[0]*0.
for i in range(n):
for j in range(n):
grid_posterior[i][j] = posterior([par_x[i][j],par_y[i][j]])
ix,iy = np.unravel_index(grid_posterior.argmax(), grid_posterior.shape)
gridmaxx = par_x[ix,iy]
gridmaxy = par_y[ix,iy]
#print gridmaxx, gridmaxy
pl.figure()
ax = pl.gca()
#xlabel = ax.get_xlabel()
#ylabel = ax.get_ylabel()
pl.title(title, fontsize=fontsize)
fig = pl.gcf()
xlabel = ax.set_xlabel(xlabel,fontsize=fontsize)
ylabel = ax.set_ylabel(ylabel,fontsize=fontsize)
#pl.xlabel(xlabel,fontsize=fontsize)
#pl.ylabel(ylabel,fontsize=fontsize)
if plot_samples:
pl.plot(x,y,'o',ms=1, mfc=samplescolor, mec=samplescolor)
for i,(inx, iny) in enumerate(zip(input_x,input_y)):
if input_label is not None:
label = input_label[i]
else:
label = 'string'
pl.plot(inx,iny,'o',mew=0.1,ms=5,color=input_color[i],label=label,zorder=4)
pl.plot(gridmaxx,gridmaxy,'x',mew=3,ms=5,color='k')
if plot_contours:
percentage_integral = np.array([0.95,0.68,0.])
contours = 0.* percentage_integral
num_epsilon_steps = 1000.
epsilon = grid_posterior.max()/num_epsilon_steps
epsilon_marks = np.arange(num_epsilon_steps + 1)
posterior_marks = grid_posterior.max() - epsilon_marks * epsilon
posterior_norm = grid_posterior.sum()
for j in np.arange(len(percentage_integral)):
for i in epsilon_marks:
posterior_integral = grid_posterior[np.where(grid_posterior>posterior_marks[i])].sum()/posterior_norm
if posterior_integral > percentage_integral[j]:
break
contours[j] = posterior_marks[i]
contours[-1]=grid_posterior.max()
pl.contour(par_x, par_y, grid_posterior, contours, linewidths=contour_lw,colors='k',zorder=3)
if fill_contours:
pl.contourf(par_x, par_y, grid_posterior, contours,colors=contour_colors,alpha=alpha,zorder=2)
pl.xlim(xmin=xmin,xmax=xmax)
pl.ylim(ymin=ymin,ymax=ymax)
if show_legend:
pl.legend(prop={'size':20},numpoints=1,loc='upper left')
if savefile is None:
return par_x, par_y, grid_posterior, contours
else:
pl.savefig(savefile, bbox_extra_artists=[xlabel, ylabel], bbox_inches='tight')
def plot_chains(mcmc, dists=None, ax=None, fig=None, multi_chain=False):
from scipy.stats import gaussian_kde as kde
from scipy.stats.mstats import mquantiles
tally = mcmc['tally']
# mcmc = mcmc0
chain = mcmc['chain'][tally:, :]
guesses = mcmc['guesses'][tally:, :]
titles = mcmc['names']
titles = [['$\Omega$', '$\phi$', r'$\rho$', '$f_1$', '$f_2$', '$f_3$'][i] for i in mcmc['active_params']]
l = len(mcmc['chain']) - tally
if fig:
axs = np.array(fig.get_axes()).reshape(len(titles), 2)
else:
fig, axs = plt.subplots(len(titles), 2, figsize=(18, 21))
axs[0, 0].set_title("Chain")
axs[0, 1].set_title("Distribution")
# axs[0, 2].set_title("Gelman Rubin")
for i, name in enumerate(titles):
if multi_chain:
ch = chain[:, i]
else:
ch = guesses[:, i]
# LEFT - Chains
# Limits
if dists:
a, b = dists[mcmc['active_params'][i]].args
a, b = a, a + b
else:
a, b = min(ch), max(ch)
qs = mquantiles(chain[:,i], [0.025,0.5,0.975])
print (qs)
a *= 0.9
b *= 1.1
ax = axs[i, 0]
if multi_chain:
my_alpha = 1 if mcmc['active'] else 0.6
ax.plot(tally + np.arange(l), chain[:, i], label=mcmc['name'], alpha=my_alpha,
zorder=5 + 1 / (10 * my_alpha)) # , label = 'chain {}'.format(j))
else:
ax.plot(tally + np.arange(l), guesses[:, i],
color='steelblue', ls='--', alpha=0.5, label="Proposed")
ax.plot(tally + np.arange(l), chain[:, i],
color='red', label="Accepted") # , label = 'chain {}'.format(j))
lims = mcmc['tally'], len(mcmc['chain'])
ax.hlines(qs,*lims,lw=1,linestyles='--', zorder=10)
ax.set_xlim(*lims)
ax.set_yticks(qs)
ax.legend()
# MIDDLE - Distribution
try:
if mcmc['active']:
density = kde(chain[:, i])
xs = np.linspace(a, b, 50)
axs[i, 1].plot(xs, density(xs))
except:
print("{} Can't KDE yet on {}".format(mcmc['name'], name))
axs[i,1].set_ylabel(name, rotation=0, fontsize=30,)
c, d = axs[i, 1].get_xlim()
a = min(a, c)
b = max(b, d)
axs[i, 1].set_xlim(a, b)
# RIGHT - Gelman Rubin
# if plot_gr:
# axs[i, 2].plot(mcmc['gelman_rubin'][2:, i], color='green')
# axs[i, 2].hlines(1, 0, len(mcmc['gelman_rubin']))
# axs[i, 2].hlines(1.1, 0, len(mcmc['gelman_rubin']), linestyles='--')
plt.tight_layout()
return fig, axs
def plot_2d_posterior(x, y,xlabel='', ylabel='',
input_x=None, input_y=None, input_color='red', fontsize=22,
contour_colors=('SkyBlue','Aqua'), alpha=0.7,
xmin=None, xmax=None, ymin=None, ymax=None,
title='', plot_samples=False, samples_color='gray',
contour_lw=2, savefile=None, plot_contours=True,
show_legend=False):
"""
The main chains plotting routine, for visualizing 2d posteriors...
"""
n = 100
points = np.array([x,y])
posterior = kde(points)
if xmin is None:
xmin=x.min()
if xmax is None:
xmax=x.max()
if ymin is None:
ymin=y.min()
if ymax is None:
ymax=y.max()
step_x = ( xmax - xmin ) / n
step_y = ( ymax - ymin ) / n
grid_pars = np.mgrid[0:n,0:n]
par_x = grid_pars[0]*step_x + xmin
par_y = grid_pars[1]*step_y + ymin
grid_posterior = grid_pars[0]*0.
for i in range(n):
for j in range(n):
grid_posterior[i][j] = posterior([par_x[i][j],par_y[i][j]])
pl.figure()
ax = pl.gca()
pl.title(title, fontsize=fontsize)
fig = pl.gcf()
xlabel = ax.set_xlabel(xlabel,fontsize=fontsize)
ylabel = ax.set_ylabel(ylabel,fontsize=fontsize)
if plot_samples:
pl.plot(x,y,'o',ms=1, mfc=samplescolor, mec=samplescolor)
pl.plot(input_x,input_y,'x',mew=3,ms=15,color=input_color,label='input',zorder=4)
if plot_contours:
percentage_integral = np.array([0.95,0.68,0.])
contours = 0.* percentage_integral
num_epsilon_steps = 1000.
epsilon = grid_posterior.max()/num_epsilon_steps
epsilon_marks = np.arange(num_epsilon_steps + 1)
posterior_marks = grid_posterior.max() - epsilon_marks * epsilon
posterior_norm = grid_posterior.sum()
for j in np.arange(len(percentage_integral)):
for i in epsilon_marks:
posterior_integral = grid_posterior[np.where(grid_posterior>posterior_marks[i])].sum()/posterior_norm
if posterior_integral > percentage_integral[j]:
break
contours[j] = posterior_marks[i]
contours[-1]=grid_posterior.max()
pl.contour(par_x, par_y, grid_posterior, contours, linewidths=contour_lw,colors='k',zorder=3)
pl.contourf(par_x, par_y, grid_posterior, contours,colors=contour_colors,alpha=alpha,zorder=2)
pl.xlim(xmin=xmin,xmax=xmax)
pl.ylim(ymin=ymin,ymax=ymax)
if show_legend:
pl.legend(prop={'size':20},numpoints=1)
if savefile is None:
return par_x, par_y, grid_posterior, contours
else:
pl.savefig(savefile, bbox_extra_artists=[xlabel, ylabel], bbox_inches='tight')
levels2 = [1., 3, 5., 7, 10., 15.]
#~
#~ levels2 = levels[::2]
cax2 = ax.contour(X, Y, F, colors="black", levels=levels2, linewidths=1.0)
ax.clabel(cax2, fontsize=12, inline=1, fmt="%i")
#~ fig.colorbar(cax)
X, Y = np.mgrid[xmin:xmax:100j, ymin:ymax:100j]
positions = np.vstack([X.ravel(), Y.ravel()])
pdf_pr1 = np.reshape(
kde(np.vstack([x_prw, dr_kink_prw]))(positions).T, X.shape)
pdf_pr2 = np.reshape(
kde(np.vstack([x_prwr, dr_kink_prwr]))(positions).T, X.shape)
pdf_pr3 = np.reshape(
kde(np.vstack([x_prwr2, dr_kink_prwr2]))(positions).T, X.shape)
pdf_pts1 = np.reshape(
kde(np.vstack([x_ptsw, dr_kink_ptsw]))(positions).T, X.shape)
pdf_pts2 = np.reshape(
kde(np.vstack([x_ptswr, dr_kink_ptswr]))(positions).T, X.shape)
pdf_pts3 = np.reshape(
kde(np.vstack([x_ptswr2, dr_kink_ptswr2]))(positions).T, X.shape)
pdf_ppa1 = np.reshape(
kde(np.vstack([x_ppaw, dr_kink_ppaw]))(positions).T, X.shape)
###########################################
################ Table 2 ##################
###########################################
print('Table 2')
print('Lipschitz bound under certain n and J values')
print(Lip_Bound_matrix)
##### Generate data for the left plot of Fig 1 #####
# Build $\pi_D^Q$ and $\pi_D^{Q,n}$, use 10,000 samples
N_kde = int(1E4)
N_mc = int(1E4)
np.random.seed(123456)
initial_sample = np.random.normal(size=N_kde)
pfprior_sample = Qexact(initial_sample)
pfprior_dens = kde(pfprior_sample)
def pfprior_dens_n(n, x):
pfprior_sample_n = Q(n, initial_sample)
pdf = kde(pfprior_sample_n)
return pdf(x)
error_r_D = np.zeros((5, 5))
np.random.seed(123456)
qsample = np.random.uniform(1, 4, N_mc)
for i in range(5):
for j in range(5):
error_r_D[i, j] = (np.mean(
(np.abs(pfprior_dens(qsample) -
# coding:utf-8 import numpy as np import matplotlib.pyplot as plt from scipy.stats import gaussian_kde as kde from scipy.stats import norm N = 1000 sigma = 0.5 sample = np.r_[np.random.normal(-1, sigma, N) , np.random.normal(1, sigma, N) ] x = np.linspace(-5, 5, 1000) y = kde(sample).evaluate(x) z = norm(-1, sigma).pdf(x) + norm(1, sigma).pdf(x) z = z / np.trapz(z, x) plt.plot(x, y, label="kde") plt.plot(x, z, label="sampling distribution") plt.hist(sample, normed=True, bins=100) plt.legend() plt.show()
def pfprior_dens_n(n, x):
pfprior_sample_n = Q(n, initial_sample)
pdf = kde(pfprior_sample_n)
return pdf(x)
def main(args):
"""
Main entrypoint for example-generation
"""
args = parse_args(args)
setup_logging(args.loglevel)
np.random.seed(args.seed)
# example = args.example
# num_trials = args.num_trials
# fsize = args.fsize
# linewidth = args.linewidth
# seed = args.seed
# inputdim = args.input_dim
# save = args.save
# alt = args.alt
# bayes = args.bayes
# prefix = args.prefix
# dist = args.dist
fdir = "figures/comparison"
check_dir(fdir)
presentation = False
save = True
if not presentation:
plt.rcParams["mathtext.fontset"] = "stix"
plt.rcParams["font.family"] = "STIXGeneral"
# number of samples from initial and observed mean (mu) and st. dev (sigma)
N, mu, sigma = int(1e3), 0.25, 0.1
lam = np.random.uniform(low=-1, high=1, size=N)
# Evaluate the QoI map on this initial sample set to form a predicted data
qvals_predict = QoI(lam, 5) # Evaluate lam^5 samples
# Estimate the push-forward density for the QoI
pi_predict = kde(qvals_predict)
# Compute more observations for use in BIP
tick_fsize = 28
legend_fsize = 24
for num_data in [1, 5, 10, 20]:
np.random.seed(
123456
) # Just for reproducibility, you can comment out if you want.
data = norm.rvs(loc=mu, scale=sigma**2, size=num_data)
# We will estimate the observed distribution using a parametric estimate to keep
# the assumptions involved as similar as possible between the BIP and the SIP
# So, we will assume the sigma is known but that the mean mu is unknown and estimated
# from data to fit a Gaussian distribution
mu_est = np.mean(data)
r_approx = np.divide(norm.pdf(qvals_predict, loc=mu_est, scale=sigma),
pi_predict(qvals_predict))
# Use r to compute weighted KDE approximating the updated density
update_kde = kde(lam, weights=r_approx)
# Construct estimated push-forward of this updated density
pf_update_kde = kde(qvals_predict, weights=r_approx)
likelihood_vals = np.zeros(N)
for i in range(N):
likelihood_vals[i] = data_likelihood(qvals_predict[i], data,
num_data, sigma)
# compute normalizing constants
C_nonlinear = np.mean(likelihood_vals)
data_like_normalized = likelihood_vals / C_nonlinear
posterior_kde = kde(lam, weights=data_like_normalized)
# Construct push-forward of statistical Bayesian posterior
pf_posterior_kde = kde(qvals_predict, weights=data_like_normalized)
# Plot the initial, updated, and posterior densities
fig, ax = plt.subplots(figsize=(10, 10))
lam_plot = np.linspace(-1, 1, num=1000)
ax.plot(
lam_plot,
uniform.pdf(lam_plot, loc=-1, scale=2),
"b--",
linewidth=4,
label="Initial/Prior",
)
ax.plot(lam_plot,
update_kde(lam_plot),
"k-.",
linewidth=4,
label="Update")
ax.plot(lam_plot,
posterior_kde(lam_plot),
"g:",
linewidth=4,
label="Posterior")
ax.set_xlim([-1, 1])
if num_data > 1:
plt.annotate(f"$N={num_data}$", (-0.75, 5),
fontsize=legend_fsize * 1.5)
ax.set_ylim([0, 28]) # fix axis height for comparisons
# else:
# ax.set_ylim([0, 5])
ax.tick_params(axis="x", labelsize=tick_fsize)
ax.tick_params(axis="y", labelsize=tick_fsize)
ax.set_xlabel("$\\Lambda$", fontsize=1.25 * tick_fsize)
ax.legend(fontsize=legend_fsize, loc="upper left")
if save:
fig.savefig(f"{fdir}/bip-vs-sip-{num_data}.png",
bbox_inches="tight")
plt.close(fig)
# plt.show()
# Plot the push-forward of the initial, observed density,
# and push-forward of pullback and stats posterior
fig, ax = plt.subplots(figsize=(10, 10))
qplot = np.linspace(-1, 1, num=1000)
ax.plot(
qplot,
norm.pdf(qplot, loc=mu, scale=sigma),
"r-",
linewidth=6,
label="$N(0.25, 0.1^2)$",
)
ax.plot(qplot,
pi_predict(qplot),
"b-.",
linewidth=4,
label="PF of Initial")
ax.plot(qplot,
pf_update_kde(qplot),
"k--",
linewidth=4,
label="PF of Update")
ax.plot(qplot,
pf_posterior_kde(qplot),
"g:",
linewidth=4,
label="PF of Posterior")
ax.set_xlim([-1, 1])
if num_data > 1:
plt.annotate(f"$N={num_data}$", (-0.75, 5),
fontsize=legend_fsize * 1.5)
ax.set_ylim([0, 28]) # fix axis height for comparisons
# else:
# ax.set_ylim([0, 5])
ax.tick_params(axis="x", labelsize=tick_fsize)
ax.tick_params(axis="y", labelsize=tick_fsize)
ax.set_xlabel("$\\mathcal{D}$", fontsize=1.25 * tick_fsize)
ax.legend(fontsize=legend_fsize, loc="upper left")
if save:
fig.savefig(f"{fdir}/bip-vs-sip-pf-{num_data}.png",
bbox_inches="tight")
plt.close(fig)
def _get_uni_kde(data, bw_method="silverman"):
return kde(data, bw_method=bw_method)
ax = axes[i]
ax.spines['left'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
ax.spines['bottom'].set_edgecolor('#444444')
ax.spines['bottom'].set_linewidth(2.0)
ax.spines['bottom'].set_zorder(3)
ax.get_yaxis().set_ticks([])
ax.get_yaxis().set_ticklabels([])
cv = data_i[::nskip,cv_index]
cv_pdf = kde(cv)
yy = cv_pdf(xx)
density_max = np.max([density_max,np.max(yy)])
if i==0:
ax.plot(xx,yy,color='orangered',linewidth=2.0,label=u"""$P_1$""")
else:
ax.plot(xx,yy,color='orangered',linewidth=2.0)
ax.fill_between(xx,yy,facecolor='orange',alpha=0.7)
cv = data_i[::nskip,cv_index+1]
cv_pdf = kde(cv)
def kde(self,
x=None,
y=None,
bw_method=None,
alpha=None,
ax=None,
interactive=True,
width=450,
height=300,
**kwds):
"""Kernel Density Estimate plot for DataFrame data
>>> dataframe.vgplot.kde() # doctest: +SKIP
Parameters
----------
x : string, optional
the column to use as the x-axis variable. If not specified, the
index will be used.
y : string, optional
the column to use as the y-axis variable. If not specified, all
columns (except x if specified) will be used.
bw_method : str, scalar or callable, optional
The method used to calculate the estimator bandwidth. This can be
'scott', 'silverman', a scalar constant or a callable.
See `scipy.stats.gaussian_kde` for more details.
alpha : float, optional
transparency level, 0 <= alpha <= 1
ax : Axes, optional
If specified, add the plot as a layer to the given axis
interactive : bool, optional
if True (default) then produce an interactive plot
width : int, optional
the width of the plot in pixels
height : int, optional
the height of the plot in pixels
Returns
-------
axes : pdvega.Axes
The vega-lite plot
"""
from scipy.stats import gaussian_kde as kde
if x is not None:
raise NotImplementedError('"x" argument to df.vgplot.kde()')
if y is not None:
df = self._data[y].to_frame()
else:
df = self._data
tmin, tmax = df.min().min(), df.max().max()
trange = tmax - tmin
t = np.linspace(tmin - 0.5 * trange, tmax + 0.5 * trange, 1000)
kde_df = pd.DataFrame(
{col: kde(df[col], bw_method=bw_method).evaluate(t)
for col in df},
index=t)
kde_df.index.name = ' '
f = FramePlotMethods(kde_df)
return f.line(value_name='Density',
alpha=alpha,
ax=ax,
interactive=interactive,
width=width,
height=height,
**kwds)
