def plot(self, logscale=True, with_validation=True, xlim=None, ylim=None):
"""
Plots the loss history.
:param logscale: if True, logarithmic scale is used
:param with_validation: if True, validation set loss is plotted
:param ylim: limits of the y-axis
"""
if 'figsize' in dir(plt):
plt.figsize(10, 5)
plt.hold(True)
try:
if logscale:
plt.yscale('log')
# plt.xscale('log')
plt.plot(self.history[0], self.history[1], 'b')
if with_validation:
plt.plot(self.history[0], self.history[2], 'c')
plt.plot(self.history[0], self.history[3], 'r')
except ValueError:
# catches: ValueError: Data has no positive values, and therefore can not be log-scaled.
# when no data is present or we only have NaNs
pass
if xlim is not None:
plt.xlim(xlim)
if ylim is not None:
plt.ylim(ylim)
yl = plt.ylim()
if with_validation and self.best_iter is not None:
plt.vlines(self.best_iter, yl[0], yl[1])
plt.xlabel('iteration')
plt.ylabel('loss')
if with_validation:
plt.legend(['training', 'validation', 'test'])
else:
plt.legend(['training', 'test'])
def plot(self, logscale=True, with_validation=True, xlim=None, ylim=None):
"""
Plots the loss history.
:param logscale: if True, logarithmic scale is used
:param with_validation: if True, validation set loss is plotted
:param ylim: limits of the y-axis
"""
if 'figsize' in dir(plt):
plt.figsize(10, 5)
plt.hold(True)
try:
if logscale:
plt.yscale('log')
# plt.xscale('log')
plt.plot(self.history[0], self.history[1], 'b')
if with_validation:
plt.plot(self.history[0], self.history[2], 'c')
plt.plot(self.history[0], self.history[3], 'r')
except ValueError:
# catches: ValueError: Data has no positive values, and therefore can not be log-scaled.
# when no data is present or we only have NaNs
pass
if xlim is not None:
plt.xlim(xlim)
if ylim is not None:
plt.ylim(ylim)
yl = plt.ylim()
if with_validation and self.best_iter is not None:
plt.vlines(self.best_iter, yl[0], yl[1])
plt.xlabel('iteration')
plt.ylabel('loss')
if with_validation:
plt.legend(['training', 'validation', 'test'])
else:
plt.legend(['training', 'test'])
def plot_lcd_tf_subplots(sub_list, coa, max_x, lcd_y=0.15, tf_y = 300, min_lc= 0):
from numpy import ceil
import matplotlib.pyplot as plt
cpool = ["#1F78B4", "#E31A1C"]
lcd_list = cluster_lcd_list(sub_list, coa)
tf_list = cluster_tf_list(sub_list, coa)
cols = 3
rows = ceil(len(lcd_list)/3.0)
height = rows * 2.2
width = cols * 4
i = 1
plt.figsize(width, height)
while i <= len(lcd_list):
lcd = lcd_list[i-1];
tf = tf_list[i-1];
lc_ax = plt.subplot(rows, cols, i);
lc_ax.plot(lcd[0], lcd[1], color=cpool[1], lw=1);
lc_ax.set_ylim(min_lc, lcd_y);
lc_ax.text(max_x*0.9, lcd_y*0.85, '%d' % i, va='top', ha='right')
tf_ax = lc_ax.twinx()
tf_ax.plot(tf[0], tf[1], color=cpool[0], alpha=0.6)
tf_ax.set_ylim(-100, tf_y)
lc_ax.set_xlim(0, max_x);
tf_ax.set_yticks([])
i = i+1;
plt.tight_layout(pad=2);
plt.subplots_adjust(top=0.96);
return plt.gcf();
def plot_corr(df, size=11):
#dataPlot = sns.load_dataset(df)
cols = [
'SalePrice', 'OverallQual', 'GrLivArea', 'GarageCars', 'TotalBsmtSF',
'FullBath', 'YearBuilt'
]
sns.pairplot(df[cols])
plt.figsize(12, 12)
plt.set_cmap('YlOrRd')
#plt.subplots(figsize=(50,100))
plt.show()
def plotAIMgraphs():
plt.figsize(40, 10)
plt.rcParams.update({'font.size': 18})
fig = plt.figure()
patches_list = []
label_list = []
for (mousetype_i, mousetype) in enumerate(["CP101", "CP73"]):
s = pd_covar.select(lambda x: pd_covar.ix[x, "MouseType"] == mousetype
and pd_covar.ix[x, "LesionType"] == "6-OHDA")
ax1 = fig.add_subplot(1, 3, mousetype_i + 1)
ax1.patch.set_facecolor((0.9, 0.9, 0.9))
ax1.grid(which='both')
plt.title(mousetype)
plt.ylabel("AIM score")
plt.xlabel("Day")
plt.ylim((-1, 50))
plt.xlim((-0.5, 4.5))
drugtreat_group_color = {
"Acute high levodopa": "green",
"Chronic saline": "grey",
"Acute saline": "yellow",
"Chronic high levodopa": "red",
"Chronic low levodopa": "blue"
}
pd_groups = s.groupby(["MouseType", "LesionType", "DrugTreat"])
for (mouse, lesion, drug) in pd_groups.groups.keys():
data = s.ix[
pd_groups.groups[(mouse, lesion, drug)],
["Day1AIM", "Day3AIM", "Day4AIM", "Day5AIM", "Day8AIM"
]].transpose()
r = np.repeat(range(len(data.index)), len(data.columns)).reshape(
(len(data.index), -1))
#xs = [range(len(data))
r = r + random.random(shape(r)) * 0.15 - 0.075
ax1.plot(r,
data,
"o-",
color=drugtreat_group_color[drug],
lw=1.5,
ms=8,
alpha=0.5,
label=lesion + " " + drug)
plt.xticks(range(0, 5), ["1", "3", "4", "5", "8"])
patches, labels = ax1.get_legend_handles_labels()
patches_list.append(patches[-1])
label_list.append(labels[-1])
ax_l = plt.subplot(1, 3, 3)
ax_l.set_frame_on(False)
ax_l.set_xticks([])
ax_l.set_yticks([])
plt.legend(patches_list[1:4], label_list[1:4], loc="upper left")
def plot_subplots(data_list, id_list, max_x):
from numpy import ceil
import matplotlib.pyplot as plt
cols = 4
rows = ceil(len(data_list)/4.0)
height = rows * 2.2
width = cols * 3
i = 1
plt.figsize(width, height)
while i <= len(data_list):
data = data_list[i-1];
plt.subplot(rows, cols, i);
plt.plot(data[0], data[1]);
plt.xlim(0, max_x);
i = i+1;
plt.tight_layout(pad=2);
plt.subplots_adjust(top=0.96);
return plt.gcf();
def pretty_plot(x, y, xlabel='', ylabel='', title='', xticks=None):
'''
Pretty plot a list.
scatter_fn -- function to select a single point to highlight
e.g., `scatter_fn=np.max` will highlight the max point of a curve
'''
plt.figsize(24, 18)
plt.grid(b=True, which='major', linestyle='--')
plt.plot(x, y, color="#348ABD", lw=3)
plt.fill_between(x, y, alpha=.2, facecolor=["#348ABD"])
plt.scatter(np.argmax(y), np.max(y), s=140, c="#348ABD")
plt.xlim(np.min(x) - .5, np.max(x) + .5)
plt.ylim(np.min(y) - .5, max(y) + .5)
plt.xlabel(xlabel)
plt.ylabel(ylabel)
if xticks:
plt.xticks(x, xticks, rotation=90)
plt.title(title)
def pretty_plot(x, y, xlabel='', ylabel='', title='', xticks=None):
'''
Pretty plot a list.
scatter_fn -- function to select a single point to highlight
e.g., `scatter_fn=np.max` will highlight the max point of a curve
'''
plt.figsize( 24, 18 )
plt.grid(b=True, which='major', linestyle='--')
plt.plot( x, y, color = "#348ABD", lw = 3 )
plt.fill_between( x, y, alpha = .2, facecolor = ["#348ABD"])
plt.scatter( np.argmax(y), np.max(y), s = 140, c ="#348ABD" )
plt.xlim(np.min(x)-.5, np.max(x)+.5)
plt.ylim(np.min(y)-.5, max(y)+.5)
plt.xlabel(xlabel)
plt.ylabel(ylabel)
if xticks:
plt.xticks(x, xticks, rotation=90)
plt.title(title);
def plot_lcd_subplots(data_list, coa, max_x=100, max_y=0.2):
from numpy import ceil
import matplotlib.pyplot as plt
data_list = cluster_lcd_list(data_list, coa)
cols = 4
rows = ceil(len(data_list)/4.0)
height = rows * 2.2
width = cols * 3
i = 1
plt.figsize(width, height)
while i <= len(data_list):
data = data_list[i-1];
plt.subplot(rows, cols, i);
plt.plot(data[0], data[1]);
plt.xlim(0, max_x);
plt.ylim(0, max_y);
plt.text(max_x/1.2, max_y/1.2, '%d' % i)
i = i+1;
plt.tight_layout(pad=2);
plt.subplots_adjust(top=0.96);
return plt.gcf();
def plotAIMgraphs():
plt.figsize(40, 10)
plt.rcParams.update({'font.size': 18})
fig = plt.figure()
patches_list = []
label_list = []
for (mousetype_i, mousetype) in enumerate(["CP101", "CP73"]):
s = pd_covar.select(lambda x: pd_covar.ix[x, "MouseType"] == mousetype and pd_covar.ix[x, "LesionType"] == "6-OHDA")
ax1 = fig.add_subplot(1, 3, mousetype_i + 1)
ax1.patch.set_facecolor((0.9, 0.9, 0.9))
ax1.grid(which='both')
plt.title(mousetype)
plt.ylabel("AIM score")
plt.xlabel("Day")
plt.ylim( (-1, 50) )
plt.xlim( (-0.5, 4.5) )
drugtreat_group_color = { "Acute high levodopa" : "green",
"Chronic saline" : "grey",
"Acute saline" : "yellow",
"Chronic high levodopa" : "red",
"Chronic low levodopa" : "blue" }
pd_groups = s.groupby(["MouseType", "LesionType", "DrugTreat"])
for (mouse, lesion, drug) in pd_groups.groups.keys():
data = s.ix[pd_groups.groups[(mouse, lesion, drug)],
["Day1AIM", "Day3AIM", "Day4AIM", "Day5AIM", "Day8AIM"]].transpose()
r = np.repeat(range(len(data.index)), len(data.columns)).reshape( (len(data.index), -1) )
#xs = [range(len(data))
r = r + random.random(shape(r)) * 0.15 - 0.075
ax1.plot(r, data, "o-", color=drugtreat_group_color[drug], lw=1.5, ms=8, alpha=0.5, label=lesion + " " + drug )
plt.xticks(range(0, 5), ["1", "3", "4", "5", "8"])
patches, labels = ax1.get_legend_handles_labels()
patches_list.append(patches[-1])
label_list.append(labels[-1])
ax_l = plt.subplot(1,3,3)
ax_l.set_frame_on(False)
ax_l.set_xticks([])
ax_l.set_yticks([])
plt.legend(patches_list[1:4], label_list[1:4], loc="upper left" )
def plot_av_tf_subplots(sub_list, coa, max_x, lcd_y=0.15, tf_y = 300):
from numpy import ceil
import matplotlib.pyplot as plt
cpool = ["#1F78B4", "#E31A1C"]
lcd_list = cluster_lcd_list(sub_list, coa)
tf_list = cluster_tf_list(sub_list, coa)
cols = 3
rows = ceil(len(lcd_list)/3.0)
height = rows * 2.2
width = cols * 4
i = 1
plt.figsize(width, height)
while i <= len(lcd_list):
lcd = lcd_list[i-1];
tf = tf_list[i-1];
ax = plt.subplot(rows, cols, i);
plt.ylim(-100, lcd_y);
plt.xlim(0, max_x)
plt.text(max_x/1.2, lcd_y/1.2, '%d' % i)
plt.plot(tf[0], tf[1], color=cpool[0], alpha=0.6)
plt.plot(lcd[0], lcd[1], color=cpool[1], lw=1);
i = i+1;
plt.tight_layout(pad=2);
def plot(self, final=True, logscale=True):
self.end_time = time.time()
if 'figsize' in dir(plt):
plt.figsize(10,5)
# plt.clf()
plt.hold(True)
if logscale:
plt.yscale('log')
#plt.xscale('log')
plt.plot(self.history[0], self.history[1], 'b')
plt.plot(self.history[0], self.history[2], 'c')
plt.plot(self.history[0], self.history[3], 'r')
yl = plt.ylim()
if self.best_iter is not None:
plt.vlines(self.best_iter, yl[0], yl[1])
plt.xlabel('iteration')
plt.ylabel('loss')
plt.legend(['training', 'validation', 'test'])
if final and self.best_iter is not None:
print "best iteration: %5d best validation test loss: %9.5f best test loss: %9.5f" % \
(self.best_iter, self.best_val_loss, self.best_tst_loss)
print "training took %.2f s" % (self.end_time - self.start_time)
from mpl_toolkits.mplot3d import Axes3D
from matplotlib import cm
import matplotlib.pyplot as plt
import numpy as np
nx=41
ny=41
dataU=np.genfromtxt("solBurgersU.dat",delimiter=",")
dataV=np.genfromtxt("solBurgersV.dat",delimiter=",")
dataU=np.delete(dataU,-1,1)
dataV=np.delete(dataV,-1,1)
x=np.linspace(0,2,nx)
y=np.linspace(0,2,nx)
X,Y=np.meshgrid(x,y)
contador=0
print type(dataU)
for u,v in zip(dataU,dataV):
a=u.reshape(nx,ny)
fig=plt.figure(plt.figsize(10,10))
ax=fig.gca(projection='3d')
wire1=ax.plot_wireframe(X,Y,a[:], cmap=cm.coolwarm)
plt.savefig("iteracion" + str(contador) + ".png")
contador+=1
print(np.sum(occurrences))
# Occurrences.mean is equal to n/N.
print("What is the observed frequency in Group A? %.4f" % np.mean(occurrences))
print("Does this equal the true frequency? %s" % (np.mean(occurrences) == p_true))
#include the observations, which are Bernoulli
with model:
obs = pm.Bernoulli("obs", p, observed=occurrences)
# To be explained in chapter 3
step = pm.Metropolis()
#trace = pm.sample(18000, step=step)
trace = pm.sample(6000, step=step, tune=1000, cores=1)
burned_trace = trace[1000:]
# -- Plot posterior of unknown p_A: --
plt.figsize(12.5, 4)
plt.title("Posterior distribution of $p_A$, the true effectiveness of site A")
plt.vlines(p_true, 0, 90, linestyle="--", label="true $p_A$ (unknown)")
plt.hist(burned_trace["p"], bins=25, histtype="stepfilled", normed=True)
plt.legend();
if 2 in run_models:
# These two quantities are unknown to us.
true_p_A = 0.05
true_p_B = 0.04
# Notice the unequal sample sizes -- no problem in Bayesian analysis.
N_A = 1500
N_B = 750
# Generate some observations..
print cm
'''
[[88 3]
[ 1 48]]'''
# To check or predict a perticular values from dataset or by our own features
example_measures = np.array([[4, 2, 1, 1, 1, 2, 3, 2, 1],
[4, 2, 1, 2, 2, 2, 3, 2, 1]])
example_measures = example_measures.reshape(len(example_measures), -1)
prediction = clf.predict(example_measures)
print prediction # [2 2]
# to print Confusion matrix
import seaborn as sns
plt.figsize(figsize=(9, 9))
sns.heatmap(cm,
annot=True,
fmt=".3f",
linewidths=.5,
square=True,
cmap='Blues_r')
plt.xlabel('Predicted label')
plt.ylabel('True label')
title = 'Accuracy Score : {0}'.format(accuracy)
plt.title(title, size=14)
plt.savefig('CM_After_prediction')
plt.show()
# We average out the expected outcomes of 1000 hands played 100 times to get a good approximation of expected utility # <codecell> l = 0 for i in list1: l = l + i/1000. print l/100. # <markdowncell> # Plot the variance. See that the mean is somewhere around -25 but it varies from one set of 1000 hands to another. # <codecell> plt.figsize(6,4) plt.figure(1) plt.hist(list1, color='k', alpha = .15) plt.show() # <codecell> print '-------' print '''Summary stats for Player 1 (never bust)''' print '-------' Size, Range, Mean, variance, skewness, kurtosis = describe(list1) print 'Number of observations: ', Size print 'Mean: ', Mean print 'Min to Max: ', Range print 'Variance: ', variance print 'Standard Dev.: ', sqrt(variance)
svm_clf.fit_transform(X, y).shape
svm_clf.fit(X, y)
scores = cross_val_score(svm_clf, X, y, cv=5)
print(scores)
param_grid = {'C': 10.**np.arange(-3, 4)}
grid_search = GridSearchCV(svm_clf, param_grid=param_grid, cv=5, verbose=3)
grid_search.fit(X, y)
print(grid_search.best_params_)
print(grid_search.best_score_)
for c in grid_search.grid_scores_:
print(c.mean_validation_score)
plt.figsize(12, 6)
plt.plot([c.mean_validation_score for c in grid_search.grid_scores_],
label="validation accuracy")
plt.xticks(np.arange(6), param_grid['C'])
plt.xlabel("C")
plt.ylabel("Accuracy")
plt.legend(loc='best')
plt.show()
lr_clf = LogisticRegression(C=0.1, penalty='l1', dual=False)
lr_clf.fit(X, y)
scores = cross_val_score(lr_clf, X, y, cv=5)
print(scores)
param_grid = {'C': 10.**np.arange(-3, 4)}
grid_search = GridSearchCV(lr_clf, param_grid=param_grid, cv=5, verbose=3)
