def compplotfreqresps(jsonfilelist, fignum=444, figname="freqfigdiff", compress=1):
""" plot the errors in the freqresp for various configurations
Parameters:
-----------
jsonfilelist : list
of json files as produced by `btu.compare_freqresp`
"""
import matplotlib.cm as cm
fig = plt.figure(fignum)
ax1 = fig.add_subplot(111)
collin = np.linspace(0.2, 0.7, len(jsonfilelist))
for k, jsfstr in enumerate(jsonfilelist):
jsf = dou.load_json_dicts(jsfstr)
tmesh = jsf["tmesh"]
redinds = range(0, len(tmesh), compress)
redina = np.array(redinds)
outsig = np.array(jsf["diffsysfr"])
curline, = ax1.plot(
np.array(tmesh)[redina], outsig[redina], c=cm.CMRmap(collin[k]), linewidth=2.0, label="{0}".format(k)
)
handles, labels = ax1.get_legend_handles_labels()
ax1.legend(handles, labels, loc="lower left")
tikz_save(figname + "{0}".format(fignum) + ".tikz", figureheight="\\figureheight", figurewidth="\\figurewidth")
print "saved to " + figname + "{0}".format(fignum) + ".tikz"
ax1.semilogx()
ax1.semilogy()
fig.show()
def compplot_deviations(jsonfilelist, fignum=333, figname="figdiff", compress=10):
""" compute the deviations of the measured output from the stable
state output
"""
import matplotlib.cm as cm
fig = plt.figure(fignum)
ax1 = fig.add_subplot(111)
diffnormlist = []
collin = np.linspace(0.2, 0.7, len(jsonfilelist))
for k, jsfstr in enumerate(jsonfilelist):
jsf = dou.load_json_dicts(jsfstr)
tmesh = jsf["tmesh"]
redinds = range(0, len(tmesh), compress)
redina = np.array(redinds)
outsig = np.array(jsf["outsig"])
# two norm at the differences
outsigdiff = np.sqrt(((outsig - outsig[0, :]) * (outsig - outsig[0, :])).sum(axis=1))
diffnormlist.append((tmesh[-1] - tmesh[0]) / len(tmesh) * outsigdiff.sum())
curline, = ax1.plot(
np.array(tmesh)[redina], outsigdiff[redina], c=cm.CMRmap(collin[k]), linewidth=2.0, label="{0}".format(k)
)
handles, labels = ax1.get_legend_handles_labels()
ax1.legend(handles, labels, loc="lower right")
tikz_save(figname + "{0}".format(fignum) + ".tikz", figureheight="\\figureheight", figurewidth="\\figurewidth")
print "saved to " + figname + "{0}".format(fignum) + ".tikz"
fig.show()
print diffnormlist
print len(tmesh)
def saveTikzPng(filename, watermark=None, thesis=False, show=False):
if watermark is not None:
plt.gcf().text(0.125, 0.9, watermark, fontsize=8)
filename_png = filename + '.png'
filename_pdf = filename + '.pdf'
fig = plt.gcf()
# plt.plot()
d = os.getcwd()
figure_folder = os.path.join(d, 'figures')
tex_folder = os.path.join(d, 'tex')
if not os.path.exists(tex_folder):
os.makedirs(tex_folder)
if not os.path.exists(figure_folder):
os.makedirs(figure_folder)
tikz_save(
tex_folder + "/" + filename + '.tex',
figureheight='\\figureheight',
figurewidth='\\figurewidth'
)
if thesis == False:
fig.savefig(figure_folder + "/" + filename_png,
format='png', dpi=600, bbox_inches='tight')
else:
fig.savefig(figure_folder + "/" + filename_pdf,
format='pdf', dpi=600, bbox_inches='tight')
fig.savefig(figure_folder + "/" + filename_png,
format='png', dpi=600, bbox_inches='tight')
if show == True:
plt.show()
def perf_profile():
erreur_abs=[0,2,5,10,15,20,25] #erreur relative!
wpp1=[91.8,91.8,95.9,100,100,100,100]
wpp2=[95.9,95.9,97.9,100,100,100,100]
#wpp3=[]
#dsatur1=[93.1,93.1,93.1,96.5,96.5,100,100]
dsatur2=[97.9,97.9,97.9,100,100,100,100]
#szekeres=[89.6,89.6,93.1,96.5,96.5,100,100]
#rlf1=[93.1,93.1,93.1,96.5,96.5,100,100]
rlf2=[97.9,97.9,97.9,100,100,100,100]
plt.xlim(-0.2,15)
plt.ylim(87,102)
plt.xlabel("Relative error compared to optimal solution")
plt.ylabel("Proportion of tests solved")
plt.title("Performance profile, |V|=25")
plot1,=plt.plot(erreur_abs,wpp1,'^-',color='0.3')
plot2,=plt.plot(erreur_abs,wpp2,'s-',color='0.3')
#plot3,=plt.plot(erreur_abs,dsatur1,'o--',color='g')
plot4,=plt.plot(erreur_abs,dsatur2,'v-',color='0.3')
#plot5,=plt.plot(erreur_abs,szekeres,'v--',color='y')
#plot6,=plt.plot(erreur_abs,rlf1,'<--',color='c')
plot7,=plt.plot(erreur_abs,rlf2,'s-',color='0.3')
plt.legend([plot1,plot2,plot4,plot7],
["WP1","WP2","DSAT","RLF"],loc="lower right")
tikz_save('/Users/romainmontagne/Desktop/Projet/article/article v3/perf_profile2.tikz',
figureheight = '\\figureheight',
figurewidth = '\\figurewidth')
def save_records_plot(file_path,
ls_monitors,
name,
n_train_batches,
legend_loc="upper right"):
"""
Save a plot of a list of monitors' history.
Args:
file_path (string): the folder path where to save the plot
ls_monitors: the list of statistics to plot
name: name of file to be saved
n_train_batches: the total number of training batches
"""
lines = ["--", "-", "-.", ":"]
linecycler = cycle(lines)
plt.figure()
for m in ls_monitors:
X = [i / float(n_train_batches) for i in m.history_minibatch]
Y = m.history_value
a, b = zip(*sorted(zip(X, Y)))
plt.plot(a, b, next(linecycler), label=m.name)
plt.xlabel('Training epoch')
plt.ylabel(ls_monitors[0].type)
plt.legend(loc=legend_loc)
plt.locator_params(axis='y', nbins=7)
plt.locator_params(axis='x', nbins=10)
plt.savefig(file_path + name + ".png")
tikz_save(file_path + name + ".tikz",
figureheight='\\figureheighttik',
figurewidth='\\figurewidthtik')
def roc_curve_per_phone(far, frr, inv_phn_map, out_file=None):
far_p = far * 100
frr_p = frr * 100
fig, ax = plt.subplots()
ax.get_xaxis().set_major_formatter(mpl.ticker.PercentFormatter())
ax.get_yaxis().set_major_formatter(mpl.ticker.PercentFormatter())
plt.xlabel('FRR')
plt.ylabel('FAR')
line = np.linspace(0, 100, num=len(far))
plt.plot(line, line)
legends = []
for i in range(len(far)):
if np.any(np.isnan(far[i])):
continue
legend, = plt.plot(frr_p[i], far_p[i], label=inv_phn_map[i])
legends.append(legend)
plt.legend(ncol=2)
# plt.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0., ncol=2)
# plt.legend(legends)
if out_file:
print("Saving plot %s" % out_file)
tikz_save(out_file + ".tex")
plt.savefig(out_file + ".png")
plt.close(fig)
else:
plt.show()
def plot(tikz,datafiles,save_figures):
key = 'symmetric_3_nodes_sib_101010'
# -----------------------------------------------------------------------------------------------------------------------------------------------------------------
# plot I(x;z) versus aib iterations after optimizing every node
# load results
with open(path_results + key + os.sep + 'Results', 'rb') as f:
tmp = pickle.load(f)
print('Results loaded')
f.close()
results = tmp.get('results')
parameters = tmp.get('params')
fig0 = plt.figure(figsize=(7, 5))
ax0 = fig0.add_subplot(111)
#do whatever plot you like ;) ...
if tikz:
tikz_save(path_plots + "....tex")
if datafiles:
np.savetxt(path_plots + '....rst', list(zip([x_values], [y_values])), delimiter='\t')
if save_figures:
fig0.savefig(path_plots + '....pdf', bbox_inches='tight', dpi=300, format='pdf')
def save_movie(self, signal, name='movie', path='movies/', **kwargs):
from matplotlib2tikz import save as tikz_save
directory = path + name
if not os.path.isdir(directory):
os.makedirs(directory)
t = self.vehicles[0].signals['time']
if ('number_of_frames' in kwargs and
kwargs['number_of_frames'] <= (t.shape[1]-1)):
number_of_frames = kwargs['number_of_frames']
else:
number_of_frames = t.shape[1]-1
root = kwargs['root'] if 'root' in kwargs else None
figurewidth = kwargs['figurewidth'] if 'figurewidth' in kwargs else '8cm'
figureheight = kwargs['figureheight'] if 'figureheight' in kwargs else None
subsample = (t.shape[1]-2)/(number_of_frames-1)
kwargs['no_update'] = True
plot = self.show(signal, **kwargs)
cnt = 0
for k in range(0, t.shape[1]-1, subsample):
self.update(k, plots=plot)
if signal == 'scene':
plt.axis('off')
path = directory+'/'+name+'_'+str(cnt)+'.tikz'
if figureheight is None:
tikz_save(path, figurewidth=figurewidth)
else:
tikz_save(path, figurewidth=figurewidth, figureheight=figureheight)
self._cleanup_rubbish(path, root)
cnt += 1
def prox_comparison():
import matplotlib.pyplot as plt
import StochProxGrad as spg
import StochProxMeth as spm
import ProxMeth as pm
import ProxGrad as pg
from matplotlib2tikz import save as tikz_save
fval_pm = pm.compute_0sr1(fn, gfn, x0, l_reg = l, tau = 1 / L, batch_size = batch_size)
fval_pg = pg.proximal_gradient(fn, gfn, x0, 1 / L, l_reg = l, batch_size = batch_size)
fval_spm = spm.compute_0sr1(f, gf, x0, X, y, l_reg = l, tau = 1 / L, batch_size = batch_size)
fval_spg = spg.proximal_gradient(f, gf, x0, 1 / L, X, y, l_reg = l, batch_size = batch_size)
fval_pm.insert(0, f(x0, X, y))
fval_pg.insert(0, f(x0, X, y))
fval_spm.insert(0, f(x0, X, y))
fval_spg.insert(0, f(x0, X, y))
line1, = plt.plot(range(len(fval_pm)), fval_pm, 'r', label = '0SR1', lw = 2)
line2, = plt.plot(range(len(fval_pg)), fval_pg, 'b', label = 'PG', lw = 2)
line3, = plt.plot(range(len(fval_spm)), fval_spm, 'g', label = 'S0SR1', lw = 2)
line4, = plt.plot(range(len(fval_spg)), fval_spg, 'k', label = 'SPG', lw = 2)
#plt.xlim([0, 55])
plt.yscale('log')
#plt.ylim([1e1, 1e13])
plt.ylabel('Function Value')
plt.xlabel('Number of Iterations')
plt.legend(handles = [line1, line2, line3, line4])
tikz_save( 'StochProx_150.tikz' );
return
def save_tex(filename, **kwargs):
"""
Save figure created with script 'filename' to a tex file using
matplotlib2tikz. Output directory is figures/. You can specify a
subdirectory using the 'subdir' kwarg.
Valid kwargs: addFigureDims and anything passed to matplotlib2tikz.save().
figureheight and -width variables will be added if addFigureDims is true.
Pass a 'suffix' if you generate multiple plots from one script.
:param filename | __file__ of the plotting script
addFigureDims | bool
"""
from matplotlib2tikz import save as tikz_save
import os.path
addFigureDims = kwargs.pop("addFigureDims", False)
kwargs["figurewidth"] = '\\figurewidth' if addFigureDims else None
kwargs["figureheight"] = '\\figureheight' if addFigureDims else None
suffix = kwargs.pop("suffix", "")
subdir = kwargs.pop("subdir", "")
kwargs["tex_relative_path_to_data"] = os.path.join("figures", subdir)
if len(subdir):
subdir += os.path.sep
tikz_save(
'../figures/{}{}{}.tex'.format(
subdir,
os.path.basename(filename).split('.')[0],
suffix),
**kwargs)
def trainFullPromp(time, Q, Qdot=None, plot_likelihoods=True, **args):
args.setdefault('print_lowerbound', plot_likelihoods)
args.setdefault('print_inner_lb', False)
args.setdefault('max_iter', 10)
if 'init_promp' in args:
robot_promp = promp.FullProMP.load(args['init_promp'])
else:
robot_promp = promp.FullProMP(basis=args['new_kernel'])
if Qdot is not None:
train_summary = robot_promp.train(time, q=Q, qd=Qdot, **args)
else:
train_summary = robot_promp.train(time, q=Q, **args)
np.set_printoptions(precision=4, linewidth=200)
print "Mean Weights:\n", robot_promp.mu_w
print "Stdev Weights:\n", np.sqrt(np.diag(robot_promp.Sigma_w))
print "Noise stdev:\n", np.sqrt(np.diag(robot_promp.Sigma_y))
print "Basis function params: ", robot_promp.get_basis_pars()
if plot_likelihoods:
lhoods = train_summary['likelihoods']
#lhoods -= lhoods[0] #Plot improvement from first iteration
plt.plot(lhoods)
plt.xlabel('Iterations')
plt.ylabel('Log-Likelihood')
if args['save_lh_plot']:
tikz_save(args['save_lh_plot'],
figureheight='\\figureheight',
figurewidth='\\figurewidth')
plt.show()
stream = robot_promp.to_stream()
fout = file(args['model_fname'], 'w')
json.dump(stream, fout)
return robot_promp
def train_log_results():
batch_size = 24
a_val = load_tf_log("Final_results/validation/dotprod_shallow.csv")
b_val = load_tf_log("Final_results/validation/pyramid.csv")
plt.style.use('ggplot')
trim = min(a_val[1].shape[0], b_val[1].shape[0])
print(b_val[2])
a_dat = a_val[2] / batch_size
b_dat = b_val[2] / batch_size
train_loss_a, = plt.plot(a_val[1][:trim],
a_dat[:trim],
label='Single Dot Product')
train_loss_b, = plt.plot(b_val[1][:trim],
b_dat[:trim],
label='Cosine Similarity Pyramid')
plt.xlabel('Epochs')
plt.ylabel('Sum of Squared Differences (SSD)')
plt.legend(handles=[train_loss_a, train_loss_b])
plt.title('Effect of Cosine Similarity Pyramid (Shallow Network)')
tikz_save('../dissertation/pyramid_comparison.tex',
figureheight='10cm',
figurewidth='14cm')
plt.clf()
def save_records_plot(file_path, stats, names, legend_loc="upper right"):
"""
Save a plot of a list of monitors' history.
Args:
file_path (string): the folder path where to save the plot
ls_monitors: the list of statistics to plot
name: name of file to be saved
n_train_batches: the total number of training batches
"""
lines = ["--", "-", "-.", ":"]
linecycler = cycle(lines)
plt.figure()
for i, s in enumerate(stats):
X = range(1, len(s) + 1)
Y = s
a, b = zip(*sorted(zip(X, Y)))
plt.plot(a, b, next(linecycler), label=names[i])
plt.xlabel('Training epoch')
plt.ylabel(names[len(names) - 1])
plt.legend(loc=legend_loc)
plt.locator_params(axis='y', nbins=7)
plt.locator_params(axis='x', nbins=10)
plt.savefig(file_path + names[len(names) - 1] + ".png")
tikz_save(file_path + names[len(names) - 1] + ".tikz",
figureheight='\\figureheighttik',
figurewidth='\\figurewidthtik')
def save_plot(self, argument=None, name='plot', path='images/', **kwargs):
from matplotlib2tikz import save as tikz_save
directory = path
if not os.path.isdir(directory):
os.makedirs(directory)
plot = self.plot(argument, **kwargs)
info = plot['info']
proj_3d = False
for k, _ in enumerate(info):
for l, _ in enumerate(info[0]):
if ('projection' in info[k][l] and
info[k][l]['projection'] == '3d'):
proj_3d = True
if proj_3d:
warnings.warn('3D plotting is not supported by matplotlib2tikz. ' +
'Saving to pdf instead.')
path = directory+'/'+name+'.pdf'
plt.savefig(path, bbox_inches='tight', pad_inches=0)
else:
figurewidth = kwargs[
'figurewidth'] if 'figurewidth' in kwargs else '8cm'
figureheight = kwargs[
'figureheight'] if 'figureheight' in kwargs else None
path = directory+'/'+name+'.tikz'
if figureheight is None:
tikz_save(path, figurewidth=figurewidth)
else:
tikz_save(
path, figurewidth=figurewidth, figureheight=figureheight)
_cleanup_rubbish(path, info)
def plt_input_par(data_par, opt=1, I=1):
# Create objective space
CrSp = createCrSp()
# Unpack data
par = unpackdatax(data_par)
# Open figure
fig = plt.figure(I, figsize=(20, 20))
ax = fig.add_subplot(111)
# Scatter plot the objective space
ax.scatter(CrSp[0], CrSp[1], s=10)
# Scatter plot of received data
ax.scatter(par[0], par[1], s=40, color='red')
# ax.scatter(par[0], par[1], s=30, color='green')
# Plot settings
plt.xlabel('$x_1$', fontsize=20)
plt.ylabel('$x_2$', fontsize=20)
plt.title('Crateria space', fontsize=20)
if opt == 1:
plt.show(block=I)
elif opt == 2:
fig.savefig('TestImg/plot_input_par_' + str(I) + '.png')
plt.close(fig)
elif opt == 3:
from matplotlib2tikz import save as tikz_save
tikz_save('TikzImg/plot_input_par_' + str(I) + '.tikz',
figureheight='\\figureheight',
figurewidth='\\figurewidth')
else:
print('No plot option is given.')
def PlotTikZ(filename, plot2save, scale=None):
try:
from matplotlib2tikz import save as tikz_save
import matplotlib
except:
return
if not isinstance(plot2save, matplotlib.figure.Figure):
plot2save = plot2save.gcf()
try:
tikz_save(filename,
figure=plot2save,
show_info=False,
float_format='%.6g')
except:
tikz_save(filename, figure=plot2save, show_info=False)
texfile = open(filename, 'rU' if sys.version_info.major < 3 else 'r')
lines = []
for line in texfile:
line = line.replace('\xe2\x88\x92', '-')
if not scale is None:
line = line.replace('begin{tikzpicture}',
'begin{tikzpicture}[scale=' + str(scale) + ']')
lines.append(str(line))
texfile.close()
texfile = open(filename, 'w')
for line in lines:
texfile.write(line)
texfile.close()
def f_decorated(*args, **kwargs):
filename = kwargs.pop('filename', None)
if filename:
makedirs(os.path.dirname(filename))
ext = os.path.splitext(filename)[1][1:] if filename else None
if ext in ('pdf', 'pgf'):
# setup fonts using Latin Modern
mpl.rcParams.update({
"font.family": ["serif"],
"font.serif": ["Latin Modern Roman"],
"font.sans-serif": ["Latin Modern Sans"],
"font.monospace": ["Latin Modern Mono"]
})
elif ext == 'png':
mpl.rcParams.update({"savefig.dpi": 240})
f(*args, **kwargs)
if filename:
if ext == "tikz":
from matplotlib2tikz import save as tikz_save
tikz_save(filename,
figureheight='\\figureheight',
figurewidth='\\figurewidth')
else:
plt.savefig(filename, bbox_inches='tight')
def plot_prs_outp(str_to_json=None, tmeshkey='tmesh', sigkey='outsig',
outsig=None, tmesh=None, fignum=222, reference=None,
compress=5, tikzfile=None, tikzonly=False):
import matplotlib.pyplot as plt
if str_to_json is not None:
jsdict = load_json_dicts(str_to_json)
tmesh = jsdict[tmeshkey]
outsig = jsdict[sigkey]
else:
str_to_json = 'notspecified'
redinds = range(1, len(tmesh), compress)
redina = np.array(redinds)
fig = plt.figure(fignum)
ax1 = fig.add_subplot(111)
ax1.plot(np.array(tmesh)[redina], np.array(outsig)[redina],
color='r', linewidth=2.0)
if tikzfile is not None:
from matplotlib2tikz import save as tikz_save
tikz_save(tikzfile + '.tikz',
figureheight='\\figureheight',
figurewidth='\\figurewidth'
)
print 'tikz saved to ' + tikzfile + '.tikz'
if tikzonly:
return
else:
fig.show()
return
def save_plot(self, argument=None, name='plot', path='images/', **kwargs):
from matplotlib2tikz import save as tikz_save
directory = path
if not os.path.isdir(directory):
os.makedirs(directory)
plot = self.plot(argument, **kwargs)
info = plot['info']
proj_3d = False
for k, _ in enumerate(info):
for l, _ in enumerate(info[0]):
if ('projection' in info[k][l]
and info[k][l]['projection'] == '3d'):
proj_3d = True
if proj_3d:
warnings.warn('3D plotting is not supported by matplotlib2tikz. ' +
'Saving to pdf instead.')
path = directory + '/' + name + '.pdf'
plt.savefig(path, bbox_inches='tight', pad_inches=0)
else:
figurewidth = kwargs[
'figurewidth'] if 'figurewidth' in kwargs else '8cm'
figureheight = kwargs[
'figureheight'] if 'figureheight' in kwargs else None
path = directory + '/' + name + '.tikz'
if figureheight is None:
tikz_save(path, figurewidth=figurewidth)
else:
tikz_save(path,
figurewidth=figurewidth,
figureheight=figureheight)
_cleanup_rubbish(path, info)
def plot_param_hist(sample,
name,
prior,
bins=30,
xlim=None,
groundtruth=None,
filename=None):
plt.cla()
plt.ticklabel_format(axis='x', style='plain')
plt.axes().get_yaxis().set_ticks([])
if xlim is None:
xlim = [np.min(sample), np.max(sample)]
binwidth = (xlim[1] - xlim[0]) / bins
x = np.arange(xlim[0], xlim[1], binwidth)
posterior = np.histogram(sample, x)[0] / (binwidth * sample.shape[0])
plt.plot(0.5 * x[:-1] + 0.5 * x[1:], posterior, 'k.-')
plt.plot(x, prior(x), 'k--')
plt.xlabel(name)
if groundtruth is not None:
plt.plot(groundtruth, 0, 'r.')
plt.legend(['Posterior', 'Prior', 'Ground truth'])
else:
plt.legend(['Posterior', 'Prior'])
if filename is not None:
tikz_save(filename,
figureheight='\\figureheight',
figurewidth='\\figurewidth')
def plot_collection(values_list,
values_names,
title="",
save_tikz=False,
tikz_path=""):
plt.style.use('ggplot')
worst = np.max(values_list)
db = lambda v: 10 * np.log(v / worst)
fig, ax = plt.subplots()
i = 0
for v in values_list:
ax.plot(db(v), label=values_names[i])
i += 1
ax.legend()
#ax.title = title
plt.xlabel('Iterations')
plt.ylabel('Cost Function (log scaled)')
plt.show()
plt.grid(True)
if save_tikz:
tikz_save(tikz_path + title + '.tex')
def save_records_plot(file_path, ls_monitors, name, n_train_batches, legend_loc="upper right"):
"""
Save a plot of a list of monitors' history.
Args:
file_path (string): the folder path where to save the plot
ls_monitors: the list of statistics to plot
name: name of file to be saved
n_train_batches: the total number of training batches
"""
lines = ["--", "-", "-.",":"]
linecycler = cycle(lines)
plt.figure()
for m in ls_monitors:
X = [i/float(n_train_batches) for i in m.history_minibatch]
Y = m.history_value
a, b = zip(*sorted(zip(X, Y)))
plt.plot(a, b, next(linecycler), label=m.name)
plt.xlabel('Training epoch')
plt.ylabel(ls_monitors[0].type)
plt.legend(loc=legend_loc)
plt.locator_params(axis='y', nbins=7)
plt.locator_params(axis='x', nbins=10)
plt.savefig(file_path + name + ".png")
tikz_save(file_path + name + ".tikz", figureheight = '\\figureheighttik', figurewidth = '\\figurewidthtik')
def f_conv_plot(fval_0sr1, fval_prox_grad):
"""
"""
import matplotlib.pyplot as plt
from matplotlib2tikz import save as tikz_save
fval_0sr1.insert(0, f(x0))
fval_prox_grad.insert(0, f(x0))
fval_l_bfgs_b.insert(0, f_l_bfgs_b(x0_l_bfgs_b))
fval_l_bfgs_b.insert(0, f_l_bfgs_b(x0_l_bfgs_b))
line1, = plt.plot(range(len(fval_0sr1)), fval_0sr1, 'r', label = '0SR1', lw = 2)
line2, = plt.plot(range(len(fval_prox_grad)), fval_prox_grad, 'b', label = 'ProxGrad', lw = 2)
line3, = plt.plot(range(len(fval_l_bfgs_b)), fval_l_bfgs_b, 'g', label = 'L-BFGS-B', lw = 2)
#plt.xscale('log')
#plt.xlim([0, 1e2])
plt.yscale('log')
#plt.ylim([0, 1e5])
plt.ylabel('Function Value')
plt.xlabel('Number of Iterations')
plt.legend(handles = [line1, line2, line3])
tikz_save( 'ProxPDE.tikz' );
plt.show()
return
def plot_TPR_TNR_ACC(x_axis,
TPR,
TNR,
ACC,
xlabel="",
ylim=None,
log_scale=False):
fig = plt.figure(xlabel + " TPR, TNR, AAC")
if log_scale:
fig.gca().set_xscale("log", nonposx='clip')
plt.plot(x_axis, TPR, label="TPR")
plt.plot(x_axis, TNR, label="TNR")
plt.plot(x_axis, ACC, label="ACC")
plt.xlabel(xlabel)
plt.ylabel("P")
plt.ylim(ylim)
plt.legend()
block_print()
tikz_save("figures/" + xlabel[0:6] + '_TPR_TNR_ACC.tikz')
enable_print()
def roc_curve(far, frr, out_file=None):
"""Plot DET curve from given FAR and FRR rates
The FAR rates are expected to be increasing while the FRR rates
are expected to be decreasing.
"""
far_p = far * 100
frr_p = frr * 100
fig, ax = plt.subplots()
ax.get_xaxis().set_major_formatter(mpl.ticker.PercentFormatter())
ax.get_yaxis().set_major_formatter(mpl.ticker.PercentFormatter())
plt.xlabel('FRR')
plt.ylabel('FAR')
line = np.linspace(0, 100, num=len(far))
plt.plot(line, line)
plt.plot(frr_p, far_p)
if out_file:
print("Saving plot %s" % out_file)
tikz_save(out_file + ".tex")
plt.savefig(out_file + ".png")
plt.close(fig)
else:
plt.show()
def plot_regions(regions, xlim, ylim, tikz=False):
"""
:param regions: dictionary with name: polytope
:param xlim: x axis limits
:param ylim: y axis limits
:param tikz: generate tikz tex code
:return: figure
"""
fig = plt.figure()
ax = fig.add_subplot(111)
for name in regions.keys():
patch = matplotlib.patches.Polygon(pc.extreme(regions[name]))
lx, ux = pc.bounding_box(regions[name]) # lower and upperbounds over all dimensions
#patch = patch_ellips(Meps, pos=None, number=number)
ax.add_patch(patch)
plt.text(ux[0], ux[1], name)
#return
#ax.add_patch(patch)
plt.xlim(xlim)
plt.ylim(ylim)
if tikz:
from matplotlib2tikz import save as tikz_save
tikz_save('regions.tex', figureheight='\\figureheight', figurewidth='\\figurewidth')
# plt.tight_layout()
plt.show()
def draw_difference_pianoroll(original, predicted, name_1='Original', name_2='Predicted', show=False, save_path=''):
if original.shape!=predicted.shape:
print("Shape mismatch. Not drawing a plot.")
return
draw_matrix = original + 2 * predicted
cm = colors.ListedColormap(['white', 'blue', 'red', 'black'])
bounds=[0,1,2,3,4]
n = colors.BoundaryNorm(bounds, cm.N)
original_color = cm(1/3)
predicted_color = cm(2/3)
both_color = cm(1.0)
original_patch = mpatches.Patch(color=original_color, label=name_1)
predicted_patch = mpatches.Patch(color=predicted_color, label=name_2)
both_patch = mpatches.Patch(color=both_color, label='Notes in both songs')
plt.figure(figsize=(20.0, 10.0))
plt.title('Difference-Pitch-plot of ' + name_1 + ' and ' + name_2, fontsize=10)
plt.legend(handles=[original_patch, predicted_patch, both_patch], loc='upper right', prop={'size': 8})
plt.pcolor(draw_matrix, cmap=cm, vmin=0, vmax=3, norm=n)
if show:
plt.show()
if len(save_path) > 0:
plt.savefig(save_path)
tikz_save(save_path + ".tex", encoding='utf-8', show_info=False)
plt.close()
def visualize_classifier_results(training_ids,
test_id,
dl,
show_classifier=True):
if show_classifier:
X_train, y_train = dl.load(training_ids, iid=True)
X_test, y_test = dl.load([test_id], iid=False)
clf = xgb.XGBClassifier(learning_rate=0.1,
n_estimators=101,
max_depth=3,
min_child_weight=3,
gamma=0.3,
subsample=0.9,
colsample_bytree=0.6,
scale_pos_weight=1,
reg_alpha=0.01,
objective='binary:logistic',
nthread=data.N_JOBS,
random_state=42)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
else:
y_pred = None
wrist_oxygen, wrist_oxygen_clean, true_oxygen = dl.load_oxygen(
test_id, y_pred=y_pred, iid=False)
wrist_oxygen = wrist_oxygen[::5]
wrist_oxygen_clean = wrist_oxygen_clean[::5]
true_oxygen = true_oxygen[::5]
if show_classifier:
graph_df = pd.concat([wrist_oxygen, true_oxygen, wrist_oxygen_clean],
axis=1,
sort=True)
else:
graph_df = pd.concat([wrist_oxygen, true_oxygen], axis=1, sort=True)
assert (wrist_oxygen.shape == true_oxygen.shape)
assert (graph_df.shape[0] == wrist_oxygen.shape[0])
# plt.figure(figsize=(4 * 1.2, 3 * 1.2))
graph_df.plot.line(figsize=(4 * 1.2, 2 * 1.2))
plt.xlabel("Time (Milliseconds)")
plt.ylabel("SpO2 (%)")
plt.ylim()
plt.legend(loc='lower left')
if show_classifier:
plt.savefig(data.GRAPH_CACHE +
'classifier-{}-{}.pdf'.format(test_id, str(dl)))
tikz_save(data.LTX_CACHE +
'classifier-{}-{}.tex'.format(test_id, str(dl)))
else:
plt.savefig(data.GRAPH_CACHE +
'algos-{}-{}.pdf'.format(test_id, str(dl)))
def BarPlot(
figName,
fun0,
funOpt,
funLabel0,
funLabelOpt,
xLabel,
yLabel,
ResultsFolder,
OptName,
figType=FileTypeRendered,
Color0="#FAA43A",
ColorOpt="#5DA5DA",
figsizex=6,
figsizey=3,
width=0.5,
xspacing=0.25,
dpi=200,
xtick=True,
usetex=False,
Tikz=False,
):
plt.rc("text", usetex=usetex)
Plot = plt.figure(figsize=(figsizex, figsizey), dpi=dpi)
ax = Plot.add_subplot(111)
nf = np.size(fun0)
ind = np.arange(nf)
rects1 = ax.bar(ind + xspacing * 2.5, fun0, width, color=Color0)
rects2 = ax.bar(ind + xspacing * 2.5 + width / 2, funOpt, width, color=ColorOpt)
lgd = ax.legend(
(rects1[0], rects2[0]),
(funLabel0, funLabelOpt),
frameon=False,
prop=fontP,
bbox_to_anchor=(1.05, 1),
loc=2,
borderaxespad=0.0,
)
ax.spines["right"].set_visible(False)
ax.spines["top"].set_visible(False)
ax.yaxis.set_ticks_position("left")
ax.xaxis.set_ticks_position("bottom")
ax.xaxis.set_major_locator(ticker.MaxNLocator(integer=True))
plt.xlim(xmin=xspacing * 2, xmax=nf + width / 2 + xspacing)
plt.ylim(ymin=np.min((np.min(fun0), np.min(funOpt), 0.0)), ymax=np.max((np.max(fun0), np.max(funOpt))))
if xtick == False:
plt.tick_params(axis="x", which="both", bottom="off", labelbottom="off")
plt.xlabel(xLabel)
plt.ylabel(yLabel)
plt.tight_layout()
for ii in range(np.size(figType)):
plt.savefig(
ResultsFolder + OptName + "_" + figName + "." + figType[ii], bbox_extra_artists=(lgd,), bbox_inches="tight"
)
if Tikz == True:
tikz_save(ResultsFolder + OptName + "_" + figName + ".tikz")
plt.close()
fail = 0
return fail
def plot_outp_sig(str_to_json=None,
tmeshkey='tmesh',
sigkey='outsig',
outsig=None,
tmesh=None,
fignum=222,
reference=None,
tikzstr=None,
compress=5):
import matplotlib.pyplot as plt
if str_to_json is not None:
jsdict = load_json_dicts(str_to_json)
tmesh = jsdict[tmeshkey]
outsig = jsdict[sigkey]
else:
str_to_json = 'notspecified'
redinds = list(range(1, len(tmesh), compress))
redina = np.array(redinds)
NY = np.int(len(outsig[0]) / 2)
fig = plt.figure(fignum)
ax1 = fig.add_subplot(111)
ax1.plot(np.array(tmesh)[redina],
np.array(outsig)[redina, :NY],
color='b',
linewidth=2.0)
ax1.plot(np.array(tmesh)[redina],
np.array(outsig)[redina, NY:],
color='r',
linewidth=2.0)
try:
from matplotlib2tikz import save as tikz_save
if tikzstr is None:
tikzstr = str_to_json + '{0}'.format(fignum)
tikz_save(tikzstr + '.tikz',
figureheight='\\figureheight',
figurewidth='\\figurewidth')
print('tikz saved to ' + tikzstr + '.tikz')
haztikz = True
except ImportError:
haztikz = False
print('cannot save to tikz -- no matplotlib2tikz found')
fig.show()
if reference is not None:
fig = plt.figure(fignum + 1)
ax1 = fig.add_subplot(111)
ax1.plot(tmesh, np.array(outsig) - reference)
if haztikz:
tikz_save(str_to_json + '{0}'.format(fignum) + '_difftoref.tikz',
figureheight='\\figureheight',
figurewidth='\\figurewidth')
fig.show()
def save_or_show_plot(name1, name2='', force_show=False, export_tikz=False):
if not visualize and not force_show:
return
filename = JOB_DIR + '/' + name1 + '_id' + str(task_id) + name2
plt.savefig(filename + '.png')
if export_tikz:
tikz_save(filename + '.tex')
plt.close()
def plot_data(df, dof, t, name):
group = df.groupby(0)
mean_df = group.mean()
std_df = group.std(ddof=1)
err_df = t * (std_df / math.sqrt(dof))
mean_df.plot(yerr=err_df, capsize=1)
#plt.show()
tikz_save(name + ".tex")
def EstabilidadeLongitudinal():
# Estabilidade da asa
CL_alpha_w = Cl_alpha_w / (
1 + Cl_alpha_w /
(np.pi * AR_w)) # [rad^-1] inclinação da curva CL x alpha na asa
CL_0_w = CL_alpha_w * abs(
alpha_0) # coeficiente de sustentação com ângulo de ataque zero
Cm_0_w = Cm_ac_w + CL_0_w * (x_cg / c_w - x_ac_w / c_w) # [rad^-1]
Cm_alpha_w = CL_alpha_w * (
x_cg / c_w - x_ac_w / c_w
) # [rad^-1] inclinação da curva Cm x alpha na asa
# Estabilidade da cauda
eta = 0.9 # razão de pressões dinâmicas na cauda e na asa
E_0 = 2 * CL_0_w / (np.pi * AR_w) # [rad] ângulo de downwash
downwash_slope = 2 * CL_alpha_w / (
np.pi * AR_w) # [rad^-1] inclinação do termo de downwash x alpha
Cm_0_t = eta * VH * CL_alpha_t * (E_0 + i_w - i_t) # [rad^-1]
Cm_alpha_t = -eta * VH * CL_alpha_t * (
1 - downwash_slope) # [rad^-1] inclinação de Cm x alpha na cauda
# Estabilidade total
Cm_alpha_total = Cm_alpha_w + Cm_alpha_t
Cm_0_total = Cm_0_w + Cm_0_t
alpha = np.linspace(0, np.pi / 8.0, 2)
y_w = Cm_alpha_w * alpha + Cm_0_w
y_t = Cm_alpha_t * alpha + Cm_0_t
y_total = Cm_alpha_total * alpha + Cm_0_total
print "Cm_alpha ASA:\t%+.3f\nCm_alpha CAUDA:\t%+.3f\nCm_alpha TOTAL:\t%+.3f\t\n" % (
Cm_alpha_w, Cm_alpha_t, Cm_alpha_total)
print "Cm_0 ASA:\t%+.3f\nCm_0 CAUDA:\t%+.3f\nCm_0 TOTAL:\t%+.3f\t\n" % (
Cm_0_w, Cm_0_t, Cm_0_total)
# PLOT Cm x alpha
plt.axhline(0, color='k')
plt.plot(alpha * 180.0 / np.pi, y_w, label='wing')
plt.plot(alpha * 180.0 / np.pi, y_t, label='tail')
plt.plot(alpha * 180.0 / np.pi, y_total, label='total')
plt.legend()
plt.xlabel("alpha")
plt.ylabel("Cm")
plt.grid()
tikz_save("p2.tex")
plt.show()
if (Cm_alpha_total < 0):
print "Aeronave Estável!!"
return Cm_alpha_total
else:
print "Aeronave Instável :("
return 0
def export_plot_tikz(self):
filename = QFileDialog.getSaveFileName(self,
'Save Plot as Tikzpicture', '.',
'*tex')
fileextension = filename[1].split('*')[1]
filename = filename[0].split('.')[0]
filename = filename + '.' + fileextension
if not len(filename) == 0:
tikz_save(filename, self.plotAreaWidget.fig)
def scatter_points(x, y, labels):
# We create a scatter plot.
df = pandas.DataFrame({
'x': x[:, 0],
'y': x[:, 1],
'words': y[:],
'labels': labels
})
df.to_csv("tsne_data.csv",
sep=" ",
columns=["x", "y", "words", "labels"],
index=False)
df_win1 = points_from_window(df, xmin=5, xmax=10, ymin=5, ymax=10)
df_win2 = points_from_window(df, xmin=-15, xmax=-10, ymin=-20, ymax=-15)
df_win1.to_csv("win1.csv",
sep=" ",
columns=["x", "y", "words", "labels"],
index=False)
df_win2.to_csv("win2.csv",
sep=" ",
columns=["x", "y", "words", "labels"],
index=False)
f = plt.figure(figsize=(8, 8))
plt.axes().set_aspect('equal')
oov_i = [i for i in range(len(labels)) if labels[i] == "OOV"]
not_oov_i = [i for i in range(len(labels)) if labels[i] == "Not OOV"]
plt.scatter(x[oov_i, 0],
x[oov_i, 1],
y[oov_i],
lw=0,
s=4,
c='red',
label="OOV")
plt.scatter(x[not_oov_i, 0],
x[not_oov_i, 1],
y[not_oov_i],
lw=0,
s=4,
c='blue',
label="Not OOV")
plt.xlim(-25, 25)
plt.ylim(-25, 25)
plt.axis('on')
plt.axis('tight')
plt.legend(labels=["OOV", "Not OOV"])
# for i, txt in enumerate(y):
# ax.annotate(txt[0], (x[i, 0], x[i, 1]))
from matplotlib2tikz import save as tikz_save
tikz_save('test.tex', figure=f)
save_file = os.path.join(cfg.VIS_SAVE_DIR, 'tsne_file.png')
plt.savefig(save_file, dpi=120)
return f
def helper_save(plt_file_name):
if plt_file_name is None:
plt.show()
else:
plt.tight_layout()
plt.savefig(plt_file_name+'.pdf', format='pdf', dpi=1000)
from matplotlib2tikz import save as tikz_save
# tikz_save('../report/ex1/plots/test.tex', figureheight='4cm', figurewidth='6cm')
tikz_save(plt_file_name + ".tikz", figurewidth="\\matplotlibTotikzfigurewidth", figureheight="\\matplotlibTotikzfigureheight",strict=False)
def plot_tensorboard_scalar(
eventlogpath,
value, # Error, Accuracy, Cost
export=False):
"""plotting a scalar from tensorboard
:param eventlogpath path to event log file (include '/' at the end)
:param value: specify the value to be printed: 'Error', 'Accuracy' or 'Cost'
:param export: enable tikz export
:type eventlogpath str
:type value str
:type export: bool
"""
path_train = eventlogpath + '/graph/train/'
path_val = eventlogpath + '/graph/val/'
ea_train = event_accumulator.EventAccumulator(path_train)
ea_val = event_accumulator.EventAccumulator(path_val)
ea_train.Reload()
ea_val.Reload()
if value == 'Error':
scalar = 'CNN_Model/Accuracy/Error'
elif value == 'Accuracy':
scalar = 'CNN_Model/Accuracy/Accuracy'
elif value == 'Cost':
scalar = 'CNN_Model/Cost'
else:
raise ValueError('specified value %s not defined', value)
train_scalar = [(s.step, s.value) for s in ea_train.Scalars(scalar)]
val_scalar = [(s.step, s.value) for s in ea_val.Scalars(scalar)]
train_scalar_x = [x[0] for x in train_scalar]
train_scalar_y = [x[1] for x in train_scalar]
val_scalar_x = [x[0] for x in val_scalar]
val_scalar_y = [x[1] for x in val_scalar]
plt.plot(train_scalar_x,
train_scalar_y,
label='Training',
color=TUMgray1,
linestyle='--',
linewidth=1.5)
plt.plot(val_scalar_x,
val_scalar_y,
label='Validierung ',
color=TUMblack,
linestyle='-',
linewidth=1.5)
plt.xlabel('Iterationen')
plt.ylabel('Klassifikationsfehler' if value == 'Error' else 'Kosten')
plt.grid(True)
plt.legend()
plt.savefig(eventlogpath + '/plot_' + value + '.pdf')
if export:
tikz_save(eventlogpath + '/plot_' + value + '.tikz', encoding='utf8')
plt.close()
def plot_re_rect_occlusion(eval_args, eval_dir, scene_ids, all_test_visibs, bins = 10):
obj_id = eval_args.getint('DATA','obj_id')
top_n_eval = eval_args.getint('EVALUATION','TOP_N_EVAL')
top_n = eval_args.getint('METRIC','TOP_N')
# if top_n_eval < 1:
# return
all_angle_errs = []
for scene_id in scene_ids:
if not os.path.exists(os.path.join(eval_dir,'error=re_ntop=%s' % top_n,'errors_{:02d}.yml'.format(scene_id))):
print 'WARNING: ' + os.path.join(eval_dir,'error=re_ntop=%s' % top_n,'errors_{:02d}.yml'.format(scene_id)) + ' not found'
continue
angle_errs_dict = inout.load_yaml(os.path.join(eval_dir,'error=re_ntop=%s' % top_n,'errors_{:02d}.yml'.format(scene_id)))
all_angle_errs += [angle_e['errors'].values()[0] for angle_e in angle_errs_dict]
if len(all_angle_errs) == 0:
return
all_angle_errs = np.array(all_angle_errs)
# print all_vsd_errs
fig = plt.figure()
plt.grid()
plt.ylabel('rot err [deg]')
plt.xlabel('visibility [percent]')
# plt.axis((-0.1, 1.1, -0.1, 1.1))
# plt.xlim((0.0, 1.0))
# plt.ylim((0.0, 1.0))
total_views = len(all_angle_errs)/top_n
angle_errs_rect = np.empty((total_views,))
for view in xrange(total_views):
top_n_errors = all_angle_errs[view*top_n:(view+1)*top_n]
angle_errs_rect[view] = np.min([top_n_errors[0], 180-top_n_errors[0]])
bounds = np.linspace(0,1,bins+1)
bin_angle_errs = []
bin_count = []
for idx in xrange(bins):
bin_idcs = np.where((all_test_visibs>bounds[idx]) & (all_test_visibs<bounds[idx+1]))
# median_angle_err[idx] = np.median(angle_errs_rect[bin_idcs])
bin_angle_errs.append(angle_errs_rect[bin_idcs])
bin_count.append(len(bin_idcs[0]))
middle_bin_vis = bounds[:-1] + (bounds[1]-bounds[0])/2.
# plt.bar(middle_bin_vis,median_angle_err,0.5/bins)
plt.boxplot(bin_angle_errs, positions = middle_bin_vis, widths=0.5/bins, sym='+')
# count_str = 'bin count ' + bins * '%s '
# count_str = count_str % tuple(bin_count)
plt.title('Visibility vs Median Rectified Rotation Error' + str(bin_count))
plt.savefig(os.path.join(eval_dir,'figures','R_err_occlusion_{:02d}.png'.format(obj_id)), dpi=300)
tikz_save(os.path.join(eval_dir,'latex','R_err_occlusion.tex'), figurewidth ='0.45\\textheight', figureheight='0.45\\textheight', show_info=False)
def SingleConvPlot(
figName,
data,
dataLabel,
xLabel,
yLabel,
ResultsFolder,
OptName,
figType=FileTypeRendered,
figsizex=6,
figsizey=3,
width=0.5,
xspacing=0.25,
dpi=200,
xtick=True,
usetex=False,
Tikz=False,
labelSubscripts=True,
):
plt.rc("text", usetex=usetex)
Plot = plt.figure(figsize=(figsizex, figsizey), dpi=dpi)
ax = Plot.add_subplot(111)
jet = plt.get_cmap("jet")
cNorm = colors.Normalize(vmin=0, vmax=len(data[0]))
scalarMap = cm.ScalarMappable(norm=cNorm, cmap=jet)
for n in range(np.size(data[0])):
if labelSubscripts == True:
if usetex == True:
ax.plot(data[:, n], color=scalarMap.to_rgba(n), label=dataLabel % str(n + 1))
else:
ax.plot(data[:, n], color=scalarMap.to_rgba(n), label=dataLabel % str(n + 1))
else:
ax.plot(data[:, n], color=scalarMap.to_rgba(n), label=dataLabel)
plt.xlabel(xLabel)
plt.ylabel(yLabel)
plt.xlim(xmin=0, xmax=len(data) - 1)
ax.spines["right"].set_visible(False)
ax.spines["top"].set_visible(False)
ax.yaxis.set_ticks_position("left")
ax.xaxis.set_ticks_position("bottom")
ax.xaxis.set_major_locator(ticker.MaxNLocator(integer=True))
numColx = (np.size(data[0]) / 12) + 1
if numColx > 3:
numColx = 3
plt.yticks(size=10 + (numColx - 1) * 4)
lgd = ax.legend(bbox_to_anchor=[1.05, 1], loc=2, ncol=numColx, frameon=False, prop=fontPP)
plt.tight_layout()
for ii in range(np.size(figType)):
plt.savefig(
ResultsFolder + OptName + "_" + figName + "." + figType[ii], bbox_extra_artists=(lgd,), bbox_inches="tight"
)
if Tikz == True:
tikz_save(ResultsFolder + OptName + "_" + figName + ".tikz")
plt.close()
fail = 0
return fail
def main():
parser = argparse.ArgumentParser(description='Radiomics results')
parser.add_argument('-svm', '--svm', metavar='svm',
nargs='+', dest='svm', type=str, required=True,
help='SVM file (HDF)')
parser.add_argument('-pinfo', '--pinfo', metavar='pinfo',
nargs='+', dest='pinfo', type=str, required=True,
help='Patient Info File (txt)')
args = parser.parse_args()
if type(args.svm) is list:
args.svm = ''.join(args.svm)
if type(args.pinfo) is list:
args.pinfo = ''.join(args.pinfo)
prediction = pd.read_hdf(args.svm)
n_class = 1
label_type = prediction.keys()[0] #NOTE: Assume we want to have the first key
N_1 = len(prediction[label_type].Y_train[0])
N_2 = len(prediction[label_type].Y_test[0])
_, predictions = plot_multi_SVM(prediction, args.pinfo, label_type, show_plots=False,
key=None, n_classifiers=[n_class], alpha=0.95)
y_score = predictions[str(n_class)]['y_score']
y_test = predictions[str(n_class)]['y_test']
plot = 'default'
print(len(y_test), len(y_score))
# y_score = np.asarray(y_score)
# y_test = np.asarray(y_test)
plot = 'default'
f, fpr, tpr = plot_ROC_CIc(y_tests, y_scores, N_1, N_2)
if plot == 'default':
plot = '_' + str(n_class)
output = os.path.join(args.svm, 'roc' + plot + '.png')
f.savefig(output)
print(("ROC saved as {} !").format(output))
output = os.path.join(args.svm, 'roc' + plot + '.tex')
tikz_save(output)
# Save ROC values as JSON
output = os.path.join(args.svm, 'roc' + plot + '.csv')
with open(output, 'wb') as csv_file:
writer = csv.writer(csv_file)
writer.writerow(['FPR', 'TPR'])
for i in range(0, len(fpr)):
data = [str(fpr[i]), str(tpr[i])]
writer.writerow(data)
print(("ROC saved as {} !").format(output))
def conv_singular_values():
"""Get singular values."""
with tf.Graph().as_default() as g:
saver = tf.train.import_meta_graph(FLAGS.checkpoint_dir + FLAGS.graph,
clear_devices=True)
with tf.Session().as_default() as sess:
saver.restore(sess,
tf.train.latest_checkpoint(FLAGS.checkpoint_dir))
svd_list = list()
for op in g.get_operations():
if op.type == 'Conv2D':
print '---'
print op.name
if len(op.inputs) == 2:
feature_map, kernel = op.inputs
print feature_map.shape.as_list()
batch_size, num_channels, height, width = feature_map.shape.as_list(
)
print kernel.shape.as_list()
kernel_size_height, kernel_size_width, input_channels, output_channels = kernel.shape.as_list(
)
if ((FLAGS.filter_one_by_one == False)
or (kernel_size_height > 1)
or (kernel_size_width > 1)):
kernel_np = sess.run(kernel)
tf_np = convsv.SVD_Conv_Tensor_NP(
kernel_np, [height, width])
this_layers_svds = np.flip(
np.sort(tf_np.flatten()), 0)
svd_list.append([op.name, this_layers_svds])
sys.stdout.flush()
n = float(len(svd_list))
for i in range(len(svd_list)):
name, svds = svd_list[i]
normalized_layer_number = (0.1 + get_layer_number(name)) / (
0.2 + MAX_LAYER_NUMBER)
this_color = (1 - normalized_layer_number, normalized_layer_number,
0.1)
short_name = name.replace('resnet_model/', '')
short_name = short_name.replace('/Conv2D', '')
plt.plot(range(len(svds)), svds, label=short_name, color=this_color)
axes = plt.gca()
plt.legend(fontsize='xx-small', ncol=3)
axes.set_xlim([0, FLAGS.xlim])
plt.xlabel('Singular value rank', fontsize=20)
plt.ylabel('Singular value', fontsize=20)
png_output_name = FLAGS.plot_output_prefix + ".png"
plot_directory = os.path.dirname(png_output_name)
if not os.path.isdir(plot_directory):
os.mkdir(plot_directory)
f = open(png_output_name, 'w')
plt.savefig(f, dpi=256)
tikz_save(FLAGS.plot_output_prefix + '.tex')
def show(self, blocking=True):
if blocking:
self.plot_data()
if self.tikz_export:
tikz_save(self.tikz_file)
plt.show()
else:
self.clean_axes()
self.plot_data()
plt.pause(0.1)
def graph(datamodel, target_list, selected_metrics, name_replace_list):
xdata = []
ydata = []
for i, target in enumerate(target_list):
# Read data and initialize data model
f = open(target, 'r')
target_data = datamodel(f.read())
if i == 0:
# On first pass, create header list
first = target_data
if selected_metrics is None:
metrics = target_data.get_all_titles()
else:
metrics = target_data.get_title_list(selected_metrics)
else:
# On subsequent passes, ensure they all have the same metrics and titles
first.check_titles_match(target_data)
# Perform find/replace on filename
target_name = target
for (fstr, rstr) in name_replace_list:
target_name = target_name.replace(fstr, rstr)
xdata.append(target_name)
# Append all the data
if selected_metrics is None:
ydata.append(target_data.get_all_data())
else:
ydata.append(target_data.get_data_list(selected_metrics))
ydata = transpose(ydata)
# Create graph
fig = plt.figure()
x = range(len(xdata))
for i, metric in enumerate(metrics):
plt.plot(x, ydata[i], 'o-', lw=4.0, label=metric)
# Set temporary x-axis to text xdata
plt.xticks(x, xdata)
plt.legend(loc='lower right', bbox_to_anchor=(0.5, -0.05))
# Write to a temporary file,
filename = "graph.tmp"
tikz_save(filename)
plt.close(fig)
# get the string from that file,
f = open(filename, 'r')
graph_data = f.read()
# remove the file,
f.close()
os.remove(filename)
# then return the string
return graph_data
def graph(datamodel, target_list, selected_metrics, name_replace_list):
xdata = []
ydata = []
for i, target in enumerate(target_list):
# Read data and initialize data model
f = open(target, 'r')
target_data = datamodel(f.read())
if i == 0:
# On first pass, create header list
first = target_data
if selected_metrics is None:
metrics = target_data.get_all_titles()
else:
metrics = target_data.get_title_list(selected_metrics)
else:
# On subsequent passes, ensure they all have the same metrics and titles
first.check_titles_match(target_data)
# Perform find/replace on filename
target_name = target
for (fstr, rstr) in name_replace_list:
target_name = target_name.replace(fstr, rstr)
xdata.append(target_name)
# Append all the data
if selected_metrics is None:
ydata.append(target_data.get_all_data())
else:
ydata.append(target_data.get_data_list(selected_metrics))
ydata = transpose(ydata)
# Create graph
fig = plt.figure()
x = range(len(xdata))
for i, metric in enumerate(metrics):
plt.plot(x, ydata[i], 'o-', lw=4.0, label=metric)
# Set temporary x-axis to text xdata
plt.xticks(x, xdata)
plt.legend(loc='lower right', bbox_to_anchor=(0.5, -0.05))
# Write to a temporary file,
filename = "graph.tmp"
tikz_save(filename)
plt.close(fig)
# get the string from that file,
f = open(filename, 'r')
graph_data = f.read()
# remove the file,
f.close()
os.remove(filename)
# then return the string
return graph_data
def plot(name, qualities_mes, costs_mes, qualities_th, costs_th):
fig, axes = plt.subplots(2,1)
ax1= axes[0]
texts_mes= []
for (i, (quality, cost)) in enumerate(zip(qualities_mes, costs_mes)):
texts_mes.append(ax1.text(quality, cost, str(i), ha='center', va='center'))
#print("Measured: ", q, c_cycle)
color='tab:red'
ax1.set_ylabel("cost per cycle (µs)")
ax1.set_xlabel("quality")
ax1.scatter(qualities_mes, costs_mes, label="Measured", color=color)
ax1.tick_params(axis='y', labelcolor=color)
ax1.grid(True)
ax2 = axes[1]
texts_th = []
for (i, (quality, cost)) in enumerate(zip(qualities_th, costs_th)):
texts_th.append(ax2.text(quality, cost, str(i), ha='center', va='center'))
color = 'tab:blue'
ax2.set_ylabel("cost")
ax2.set_xlabel("quality")
ax2.scatter(qualities_th, costs_th, label="Model", color=color)
ax2.tick_params(axis='y', labelcolor=color)
ax2.grid(True)
adjust_text(texts_mes, ax=ax1)
adjust_text(texts_th, ax=ax2)
kendalltau = GraphResults("Kendall's tau")
kendalltau.costs = stats.kendalltau(costs_mes, costs_th, nan_policy='raise')
kendalltau.quality = stats.kendalltau(qualities_mes, qualities_th, nan_policy='raise')
spearmanr = GraphResults("Spearman's R")
spearmanr.costs = stats.spearmanr(costs_mes, costs_th, nan_policy='raise')
spearmanr.quality = stats.spearmanr(qualities_mes, qualities_th, nan_policy='raise')
print(kendalltau.name, " Kendal's tau: cost=", kendalltau.costs, " and quality=", kendalltau.quality)
print(spearmanr.name, " Spearman's r: cost=", spearmanr.costs, " and quality=", spearmanr.quality)
fig.tight_layout()
fig.legend()
if args.tikz:
tikz_save(name+".tex")
plt.show()
def plot_outp_sig(str_to_json=None, tmeshkey='tmesh', sigkey='outsig',
outsig=None, tmesh=None, fignum=222, reference=None,
tikzstr=None, compress=5, notikz=False):
import matplotlib.pyplot as plt
if str_to_json is not None:
jsdict = load_json_dicts(str_to_json)
tmesh = jsdict[tmeshkey]
outsig = jsdict[sigkey]
else:
str_to_json = 'notspecified'
redinds = list(range(1, len(tmesh), compress))
redina = np.array(redinds)
NY = np.int(len(outsig[0])/2)
fig = plt.figure(fignum)
ax1 = fig.add_subplot(111)
ax1.plot(np.array(tmesh)[redina], np.array(outsig)[redina, :NY],
color='b', linewidth=2.0)
ax1.plot(np.array(tmesh)[redina], np.array(outsig)[redina, NY:],
color='r', linewidth=2.0)
if notikz:
plt.show()
return
try:
from matplotlib2tikz import save as tikz_save
if tikzstr is None:
tikzstr = str_to_json + '{0}'.format(fignum)
tikz_save(tikzstr + '.tikz',
figureheight='\\figureheight',
figurewidth='\\figurewidth'
)
print('tikz saved to ' + tikzstr + '.tikz')
haztikz = True
except ImportError:
haztikz = False
print('cannot save to tikz -- no matplotlib2tikz found')
fig.show()
if reference is not None:
fig = plt.figure(fignum+1)
ax1 = fig.add_subplot(111)
ax1.plot(tmesh, np.array(outsig)-reference)
if haztikz:
tikz_save(str_to_json + '{0}'.format(fignum) + '_difftoref.tikz',
figureheight='\\figureheight',
figurewidth='\\figurewidth'
)
fig.show()
def save(self, signal, name='plot', path='images/', **kwargs):
from matplotlib2tikz import save as tikz_save
directory = path
if not os.path.isdir(directory):
os.makedirs(directory)
self.show(signal, **kwargs)
if signal == 'scene':
plt.axis('off')
figurewidth = kwargs[
'figurewidth'] if 'figurewidth' in kwargs else '8cm'
path = directory+'/'+name+'.tikz'
tikz_save(path, figurewidth=figurewidth)
self._remove_rubbish(path)
def DoubleConvPlot(
figName,
data,
dataLabel,
xLabel,
yLabel,
Color,
ResultsFolder,
OptName,
figType=FileTypeRendered,
figsizex=6,
figsizey=3,
width=0.5,
xspacing=0.25,
dpi=200,
xtick=True,
usetex=False,
Tikz=False,
):
plt.rc("text", usetex=usetex)
Plot = plt.figure(figsize=(figsizex, figsizey), dpi=dpi)
ax1 = Plot.add_subplot(111)
ax1.plot(data[0], Color[0], label=dataLabel[0])
plt.xlabel(xLabel)
plt.ylabel(yLabel[0])
ax2 = ax1.twinx()
ax2.plot(data[1], Color[1], label=dataLabel[1])
plt.ylabel(yLabel[1])
handles1, labels1 = ax1.get_legend_handles_labels()
handles2, labels2 = ax2.get_legend_handles_labels()
handles = handles1 + handles2
labels = labels1 + labels2
lgd = ax1.legend(handles, labels, bbox_to_anchor=[1.05, 1], loc=2, frameon=False, prop=fontPP)
plt.xlim(xmin=0, xmax=len(data[0]) - 1)
ax1.spines["right"].set_visible(False)
ax1.spines["top"].set_visible(False)
ax2.spines["top"].set_visible(False)
ax1.yaxis.set_ticks_position("left")
ax1.xaxis.set_ticks_position("bottom")
ax1.xaxis.set_major_locator(ticker.MaxNLocator(integer=True))
plt.tight_layout()
for ii in range(np.size(figType)):
plt.savefig(
ResultsFolder + OptName + "_" + figName + "." + figType[ii], bbox_extra_artists=(lgd,), bbox_inches="tight"
)
if Tikz == True:
tikz_save(ResultsFolder + OptName + "_" + figName + ".tikz")
plt.close()
fail = 0
return fail
def _gohome(tikzfile=None, showplot=True):
if tikzfile is not None:
try:
from matplotlib2tikz import save as tikz_save
tikz_save(tikzfile,
figureheight='\\figureheight',
figurewidth='\\figurewidth')
print 'you may want to use this command\n\\input{' +\
tikzfile + '}\n'
except ImportError:
print 'matplotlib2tikz need to export tikz filez'
if showplot:
plt.show(block=False)
def outputplot(tmesh, ycomp, ystar=None, extra=None, fignum=111, fname="notspecified"):
from matplotlib2tikz import save as tikz_save
plt.figure(fignum)
if ystar is not None:
plt.plot(tmesh, ystar, color="b", linewidth=2.0)
lines = plt.plot(tmesh, ycomp)
plt.setp(lines, color="r", linewidth=2.0)
plt.yticks([0.2, 0, -0.2])
# plt.xticks([0, 2])
tikz_save(fname + ".tikz", figureheight="\\figureheight", figurewidth="\\figurewidth", extra=extra)
plt.show(block=False)
def plot_step_resp(str_to_json=None, tmesh=None,
red_stp_rsp=None, ful_stp_rsp=None, inivout=None,
compress=20):
"""
compress : real, optional
factor of compressing for plot, defaults to 20
"""
from matplotlib2tikz import save as tikz_save
if str_to_json is not None:
jsdict = dou.load_json_dicts(str_to_json)
tmesh = np.array(jsdict['tmesh'])
red_stp_rsp = jsdict['red_stp_rsp']
ful_stp_rsp = jsdict['ful_stp_rsp']
inivout = jsdict['inivout']
else:
str_to_json = 'notspecified'
redinds = list(range(0, len(tmesh), compress))
redina = np.array(redinds)
for ccol in range(len(red_stp_rsp)):
# [0, b_mat.shape[1]-1]: # range(2): # b_mat.shape[1]):
fuloutp = np.array(ful_stp_rsp[ccol])-np.array(inivout).T
redoutp = np.array(red_stp_rsp[ccol])
outdiff = fuloutp - redoutp
NY = fuloutp.shape[1]/2
fig = plt.figure(200 + ccol)
ax1 = fig.add_subplot(131)
ax1.plot(tmesh[redina], redoutp[redina, :NY], color='b', linewidth=2.0)
ax1.plot(tmesh[redina], redoutp[redina, NY:], color='r', linewidth=2.0)
ax2 = fig.add_subplot(132)
ax2.plot(tmesh[redina], fuloutp[redina, :NY], color='b', linewidth=2.0)
ax2.plot(tmesh[redina], fuloutp[redina, NY:], color='r', linewidth=2.0)
ax3 = fig.add_subplot(133)
ax3.plot(tmesh[redina], outdiff[redina, :NY], color='b', linewidth=2.0)
ax3.plot(tmesh[redina], outdiff[redina, NY:], color='r', linewidth=2.0)
tikz_save(str_to_json + '{0}'.format(200+ccol) + '.tikz',
figureheight='\\figureheight',
figurewidth='\\figurewidth'
)
print('saved to ' + str_to_json + '{0}'.format(200+ccol) + '.tikz')
fig.show()
def plot_blocks_throughput(self, data, data_2):
nblocks = data[6]
nblocks_2 = data_2[6]
# print "data 6 ", data[6]
# print "data 7 ", data[7]
# data[7] gives minimum
# data[8] gives average
# data[9] gives variance
block_dict = data[8]
block_dict_2 = data_2[8]
self.blks_x = numpy.array(xrange(nblocks))
self.blks_x_2 = numpy.array(xrange(nblocks_2))
sorted_dict = sorted(block_dict.iteritems(), key=operator.itemgetter(0))
sorted_dict.reverse()
self.blks_times = [s[1] for s in sorted_dict]
self.blks_keys = [s[0] for s in sorted_dict]
sorted_dict_2 = sorted(block_dict_2.iteritems(), key=operator.itemgetter(0))
sorted_dict_2.reverse()
self.blks_times_2 = [s[1] for s in sorted_dict_2]
self.blks_keys_2 = [s[0] for s in sorted_dict_2]
self.fig100 = plt.figure(100 + self.fignum, figsize=(12, 10), facecolor="w")
self.fig100.clf()
self.sp101 = self.fig100.add_subplot(1, 1, 1)
self.blks_times[:] = [x / self.nitems * 1000 for x in self.blks_times]
print "SUMSUM: ", sum(self.blks_times)
self.blks_times_2[:] = [x / self.nitems * 1000 for x in self.blks_times_2]
print "self.blks_times ", self.blks_times
self.sp101.barh(self.blks_x, self.blks_times, height=width, color="r", alpha=0.85, label="2.5 MHz")
self.sp101.barh(self.blks_x_2 + width, self.blks_times_2, height=width, color="b", alpha=0.85, label="1 MHz")
# print self.blks_times
self.sp101.set_title(self.title, fontsize=22, fontweight="bold")
self.sp101.set_xlabel("Execution time per OFDM symbol (us)", fontsize=16)
self.sp101.set_yticks(self.blks_x + width)
self.sp101.set_yticklabels(self.blks_keys)
self.sp101.grid()
self.sp101.legend()
# self.sp101.set_color_cycle(['c', 'm', 'y', 'k'])
plt.draw()
tikz_save("myfile.tikz", figureheight="20cm", figurewidth="14cm")
def x_conv_plot(xval_0sr1, xval_prox_grad):
"""
"""
import matplotlib.pyplot as plt
from matplotlib2tikz import save as tikz_save
d_0sr1, d_prox_grad, d_l_bfgs_b = [], [], []
for j in range(len(xval_0sr1)):
d_0sr1.append(np.linalg.norm(xval_0sr1[j] - xval_0sr1[-1]))
for j in range(len(xval_prox_grad)):
d_prox_grad.append(np.linalg.norm(xval_prox_grad[j]-xval_prox_grad[-1]))
for j in range(len(xval_l_bfgs_b)):
d_l_bfgs_b.append(np.linalg.norm(xval_l_bfgs_b[j]-xval_l_bfgs_b[-1]))
q_0sr1, q_prox_grad, q_l_bfgs_b = [], [], []
for i in range(len(d_0sr1) - 1):
q_0sr1.append(d_0sr1[i+1] / d_0sr1[i])
for i in range(len(d_prox_grad) - 1):
q_prox_grad.append(d_prox_grad[i+1] / d_prox_grad[i])
for i in range(len(d_l_bfgs_b) - 1):
q_l_bfgs_b.append(d_l_bfgs_b[i+1] / d_l_bfgs_b[i])
line1, = plt.plot(range(len(q_0sr1)), q_0sr1, 'r', label = '0SR1', lw = 2)
line2, = plt.plot(range(len(q_prox_grad)), q_prox_grad, 'b', label = 'ProxGrad', lw = 2)
line3, = plt.plot(range(len(q_l_bfgs_b)), q_l_bfgs_b, 'g', label = 'L-BFGS-B', lw = 2)
#plt.xscale('log')
plt.xlim([0, 20])
#plt.yscale('log')
#plt.ylim([0, 1e5])
plt.ylabel('Convergence Factor')
plt.xlabel('Number of Iterations')
plt.legend(handles = [line1, line2, line3])
tikz_save( 'convPDE.tikz' );
return
def plot_outp_sig(str_to_json=None, tmeshkey='tmesh', sigkey='outsig',
outsig=None, tmesh=None, fignum=222, reference=None,
compress=5):
import matplotlib.pyplot as plt
from matplotlib2tikz import save as tikz_save
if str_to_json is not None:
jsdict = load_json_dicts(str_to_json)
tmesh = jsdict[tmeshkey]
outsig = jsdict[sigkey]
else:
str_to_json = 'notspecified'
redinds = range(1, len(tmesh), compress)
redina = np.array(redinds)
NY = len(outsig[0])/2
fig = plt.figure(fignum)
ax1 = fig.add_subplot(111)
ax1.plot(np.array(tmesh)[redina], np.array(outsig)[redina, :NY],
color='b', linewidth=2.0)
ax1.plot(np.array(tmesh)[redina], np.array(outsig)[redina, NY:],
color='r', linewidth=2.0)
tikz_save(str_to_json + '{0}'.format(fignum) + '.tikz',
figureheight='\\figureheight',
figurewidth='\\figurewidth'
)
print 'tikz saved to ' + str_to_json + '{0}'.format(fignum) + '.tikz'
fig.show()
if reference is not None:
fig = plt.figure(fignum+1)
ax1 = fig.add_subplot(111)
ax1.plot(tmesh, np.array(outsig)-reference)
tikz_save(str_to_json + '{0}'.format(fignum) + '_difftoref.tikz',
figureheight='\\figureheight',
figurewidth='\\figurewidth'
)
fig.show()
def affiche(a, b):
pl.close('all')
fig = pl.figure()
ax = fig.add_subplot(111)
ax.set_title('Taille du graphe:{} Échantillon:{} K={}, q=$\\frac'
.format(graph_size,nb_gen,k)+'{'+str(b)+'}{'+str(a+b)+'}$')
ax.set_ylabel("Proportion finale")
ax.set_xlabel("Proportion initiale")
pl.grid()
pl.plot(X/100,X/100, '--', color="gray", label="Identité", lw=2.5)
for i,(beta, name, label, color) in enumerate(corres):
for x in X:
cursor = conn.execute(
"SELECT value FROM r_spreading_{}_".format(name)+
"{}_{}_{}_{}_{}_{}_{}_{}"
.format(graph_size, beta, k, nb_gen, x, max_spread_step, a, b))
rows = cursor.fetchall()
v = 0
n = len(rows)
for row in rows:
v += json.loads(row[0])['prop_spread'][-1]
Y[i][x] = v/n
v = 0
for row in rows:
v += (Y[i][x] - json.loads(row[0])['prop_spread'][-1])**2
v = (v/n)**(1/2)
Y_high[i][x] = Y[i][x] + v
Y_low[i][x] = Y[i][x] - v
pl.plot(X/100,Y[i][1:], label="{}".format(label), lw=2.5, color=color)
# for i in range(3):
# pl.plot(X, Y_high[i][1:], 'r--', lw=2)
# pl.plot(X, Y_low[i][1:], 'r--', lw=2)
# pl.axhline(0.5, xmin=0, xmax=100, c='orange', lw=2.5)
pl.legend(loc="lower right")
tikz_save("resultats/all_finale_f_initiale_q{}.tex"
.format(int(b/(a+b)*100), beta),
figurewidth="\\textwidth",
figureheight=".33\\textheight")
def harry_plotter(sim):
fig = plt.figure()
ax = fig.add_subplot(111, autoscale_on=False)
#ani = sim.animate(fig, ax)
ax.axis('scaled')
ax.set_xlim((-10, 160))
ax.set_ylim((-10, 160))
ax.set_xlabel('East [m]')
ax.set_ylabel('North [m]')
sim.draw(ax)
# ax.grid()
# plt.tight_layout()
tikz_save('../../../latex/fig/astar-admissibility-2.tikz',
figureheight='1.5\\textwidth',
figurewidth='1.5\\textwidth')
plt.show()
def plot(self):
plt.clf()
lines = plt.plot(range(1, self.i+1), self.training_measures[:self.i])
plt.legend(iter(lines), ('accuracy', 'recall', 'precision'), loc=4)
plt.xlabel('Iteration')
plt.ylabel('Measures')
plt.axis([1, self.i+1, 0, 1])
plt.title('Training measures over time')
plt.savefig('training_measures.png')
tikz_save('training_measures.tikz', show_info=False, figurewidth='0.5\\textwidth')
plt.clf()
lines = plt.plot(range(1, self.i+1), self.dev_measures[:self.i])
plt.legend(iter(lines), ('accuracy', 'recall', 'precision'), loc=4)
plt.xlabel('Iteration')
plt.ylabel('Measures')
plt.axis([1, self.i+1, 0, 1])
plt.title('Development measures over time')
plt.savefig('dev_measures.png')
tikz_save('dev_measures.tikz', show_info=False, figurewidth='0.5\\textwidth')
plt.clf()
lines = plt.plot(range(1, self.i+1), self.testing_measures[:self.i])
plt.legend(iter(lines), ('accuracy', 'recall', 'precision'), loc=4)
plt.xlabel('Iteration')
plt.ylabel('Measures')
plt.axis([1, self.i+1, 0, 1])
plt.title('Testing measures over time')
plt.savefig('testing_measures.png')
tikz_save('testing_measures.tikz', show_info=False, figurewidth='0.5\\textwidth')
def on_epoch_end(self, epoch, logs={}):
p = model.predict(X, verbose=1)
#p = model.predict({'input': X}, verbose=1)['output']
train_acc = accuracy_score(Y[:t1], np.round(p[:t1]))
dev_acc = accuracy_score(Y[t1:t2], np.round(p[t1:t2]))
test_acc = accuracy_score(Y[t2:], np.round(p[t2:]))
print('Accuracy | train:', train_acc, 'dev:', dev_acc, 'test:', test_acc)
train_recall = (recall_score(Y[:t1], np.round(p[:t1])) + recall_score(Y[:t1], np.round(p[:t1]), pos_label=0)) / 2
dev_recall = (recall_score(Y[t1:t2], np.round(p[t1:t2])) + recall_score(Y[t1:t2], np.round(p[t1:t2]), pos_label=0)) / 2
test_recall = (recall_score(Y[t2:], np.round(p[t2:])) + recall_score(Y[t2:], np.round(p[t2:]), pos_label=0)) / 2
print('Recall | train:', train_recall, 'dev:', dev_recall, 'test:', test_recall)
self.accuracy[epoch, :] = np.array([train_acc, dev_acc, test_acc])
self.recall[epoch, :] = np.array([train_recall, dev_recall, test_recall])
plt.clf()
plt.subplot(211)
lines = plt.plot(range(1, epoch+2), self.accuracy[:epoch+1])
plt.legend(iter(lines), ('train', 'dev', 'test'), loc=4)
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.axis([1, epoch+2, 0, 1])
plt.subplot(212)
lines = plt.plot(range(1, epoch+2), self.recall[:epoch+1])
plt.legend(iter(lines), ('train', 'dev', 'test'), loc=4)
plt.xlabel('Epoch')
plt.ylabel('Average recall')
plt.axis([1, epoch+2, 0, 1])
plt.savefig('results.png')
tikz_save('results.tikz', show_info=False, figurewidth='0.5\\textwidth')
if dev_recall > self.bestrecall:
self.bestrecall = dev_recall
model.save_weights('weights.hdf5', overwrite=True)
def GeneratePlot():
getcontext().prec += 2
pn = Decimal(1)
bn = Decimal(1)
hn = Decimal(math.sqrt(pn**2 + bn ** 2))
f = Decimal(0.5)
g = Decimal(1)
s = 0
t = range(10)
v = []
for i in t:
An = Decimal(f * pn * bn)
bn = Decimal(f * hn)
pn = Decimal(g - Decimal(math.sqrt(g- bn ** 2)))
hn = Decimal(math.sqrt(pn**2 + bn**2))
s = Decimal(s+( 2**i *An))
res = 4*s
v.append(res)
print(res)
plt.xlabel(r'Iterations $[n]$')
plt.ylabel('Value of $\pi$ ')
plt.grid(True)
plt.plot(t,v,'o')
tikz_save('../Images/PISeriesPlot.tex',figurewidth="12cm",figureheight="7cm")
def main():
plt.ion()
parser = add_argparser()
args = parser.parse_args()
data = pickle.load(open(args.file, "rb"))
tests = data[1]
results = data[2]
# print results
data_2 = pickle.load(open(args.file_two, "rb"))
tests_2 = data_2[1]
results_2 = data_2[2]
nitems = args.nitems
nfigs = 0
plots = []
for t in tests:
nfigs += 1
plots.append(plotter(t, results[t["testname"]], results_2[t["testname"]], nfigs, nitems))
tikz_save("myfile.tikz", figureheight="20cm", figurewidth="14cm")
plt.show(block=True)
