def show():
blob_centers = np.array([(-5, -5), (0, 0), (5, 5)])
cluster_centers = np.array([(-1, 1), (0, 0), (1, -1)])
X, _ = make_blobs(n_samples=30,
centers=blob_centers,
shuffle=False,
random_state=1234)
colors = np.array(sns.color_palette(n_colors=3))
fig = plt.figure(figsize=figaspect(1))
ax = fig.gca()
ax.scatter(X[:, 0], X[:, 1], c='gray', alpha=0.4)
xmin, xmax = ax.get_xlim()
ymin, ymax = ax.get_ylim()
xx, yy = np.meshgrid(np.linspace(xmin, xmax, 100),
np.linspace(ymin, ymax, 100))
mesh = np.c_[xx.ravel(), yy.ravel()]
ax.scatter(mesh[:, 0],
mesh[:, 1],
c=colors[np.argmin(cdist(mesh, cluster_centers), axis=1)],
alpha=0.2,
marker='.')
ax.scatter(cluster_centers[:, 0],
cluster_centers[:, 1],
c='white',
marker='x')
ax.set(xlim=(xmin, xmax), ylim=(ymin, ymax), xticks=(), yticks=())
plt.show()
def show():
x, y = make_regression(n_samples=8,
n_features=1,
bias=50,
noise=10,
random_state=1234)
x = StandardScaler().fit_transform(x)
model = LinearRegression().fit(x, y)
a = model.coef_
b = model.intercept_
plt.figure(figsize=figaspect(1))
ax = plt.axes()
ax.axhline(0, linewidth=1, color='gray')
ax.axvline(0, linewidth=1, color='gray')
ax.scatter(x, y, color='black', label='データ')
xlim = ax.get_xlim()
ylim = ax.get_ylim()
xlim = (min(xlim[0], -3), max(xlim[1], 3))
ylim = (min(ylim[0], -3 * a + b), max(ylim[1], 3 * a + b))
ax.plot(xlim, a * xlim + b, label='回帰直線 $y=ax+b$')
ax.hlines((a + b, b), (1, xlim[0]), (2, 0),
color='gray',
linestyle='dashed')
ax.vlines(2, a + b, 2 * a + b, color='gray', linestyle='dashed')
ax.text(1.5, (a + b) * 0.9, 1, verticalalignment='top')
ax.text(2 * 1.05, 1.5 * a + b, '$a$', horizontalalignment='left')
ax.legend()
ax.set(xlabel='$x$',
xlim=xlim,
ylim=ylim,
xticks=(0, ),
yticks=(b, ),
yticklabels=('$b$', ))
ax.set_ylabel('$y$', rotation='horizontal')
plt.show()
def show():
points = np.array([[1, 1], [2, 2]])
fig = plt.figure(figsize=figaspect(1))
ax = fig.gca()
ax.axhline(0, color='black', alpha=0.2)
ax.axvline(0, color='black', alpha=0.2)
def scatter_with_annotation(ax, text, point):
ax.scatter(*point, color='gray')
ax.text(point[0], point[1] + 0.1, text, horizontalalignment='right')
scatter_with_annotation(ax, 'A', points[0])
scatter_with_annotation(ax, 'B', points[1])
ax.annotate(None,
points[0],
xytext=points[1],
arrowprops=dict(
arrowstyle="-",
color="red",
alpha=0.4,
connectionstyle="angle, angleA = 90, angleB = 0"))
ax.text(points[1, 0],
points[0, 1],
'd',
size='large',
color='red',
horizontalalignment='left',
verticalalignment='top')
ax.set(xlim=(-0.5, 3), ylim=(-0.5, 3), xticks=(0, ), yticks=(0, ))
plt.show()
def show():
_, (ax1, ax2) = plt.subplots(
1, 2, figsize=figaspect(1/3), sharex=True, sharey=True)
v1 = venn2((2, 2, 1), set_labels=('', ''),
set_colors=('lightgray', 'white'), ax=ax1)
set_label(v1)
v1.get_patch_by_id('11').set(facecolor='blue')
ax1.set(title=r'分子$=conf( A\Rightarrow B)$')
v2 = venn2((2, 2, 1), set_labels=('', ''),
set_colors=('lightgray', 'blue'), ax=ax2)
set_label(v2)
v2.get_patch_by_id('11').set(facecolor='blue')
ax2.set(title='分母$=supp( B)$')
xmin, xmax = ax2.get_xlim()
ymin, ymax = ax2.get_ylim()
width = xmax - xmin
height = ymax - ymin
margin = 0.01
padding = 0.1
ax2.text(xmin + width*padding, ymax - height*padding, '$N$')
ax2.add_patch(Rectangle((xmin + width*margin, ymin + height*margin),
width*(1 - margin*2), height*(1 - margin*2),
facecolor='lightgray', zorder=0))
plt.show()
def simulation_analysis_plot(
system: systemCls,
title: str = "",
out_path: str = None,
limits_coordinate_space: Tuple[float, float] = None,
oneD_limits_potential_system_energy: Tuple[float, float] = None,
limits_force: Tuple[float, float] = None,
twoD_number_of_bins: int = 25,
resolution_full_space=style.potential_resolution,
figsize: Tuple[float, float] = figaspect(0.25)
) -> Tuple[plt.figure, str]:
"""
This is a wrapper function for the analysis of
Parameters
----------
system
title
out_path
limits_coordinate_space
resolution_full_space
Returns
-------
"""
if (system.nDimensions == 1):
if (hasattr(system, "bias_potential") and system.bias_potential):
fig, out_path = oneD_biased_simulation_analysis_plot(
system=system,
title=title,
out_path=out_path,
limits_coordinate_space=limits_coordinate_space,
limits_potential_system_energy=
oneD_limits_potential_system_energy,
limits_force=limits_force,
resolution_full_space=resolution_full_space,
figsize=figsize)
else:
fig, out_path = oneD_simulation_analysis_plot(
system=system,
title=title,
out_path=out_path,
limits_coordinate_space=limits_coordinate_space,
limits_potential_system_energy=
oneD_limits_potential_system_energy,
limits_force=limits_force,
resolution_full_space=resolution_full_space,
figsize=figsize)
elif (system.nDimensions == 2):
fig, out_path = twoD_simulation_analysis_plot(
system=system,
title=title,
out_path=out_path,
limits_coordinate_space=limits_coordinate_space,
number_of_bins=twoD_number_of_bins,
resolution_full_space=resolution_full_space,
figsize=figsize)
return fig, out_path
def mergepmf(pmf1, pmf2, outputname, plot=True):
# merge PMFs
cv_czar, pmf_czar = np.genfromtxt(pmf1, unpack=True)
cv_amd, pmf_amd = np.genfromtxt(pmf2, unpack=True)
#cv_count, abf_count = np.genfromtxt('deca.count', unpack = True)
#abf_count_correction = calcpmf(abf_count)
#pmf_amd += abf_count_correction
f = interp1d(cv_amd, pmf_amd, fill_value='extrapolate', kind='quadratic')
pmf_amd = f(cv_czar)
pmf_total = pmf_amd + pmf_czar
pmf_total_min = np.min(pmf_total)
pmf_total = pmf_total - pmf_total_min
output_pmf = outputname + '.pmf'
headerLines = readHeaderString(pmf1)
with open(output_pmf, 'w') as fp_out:
for line in headerLines:
fp_out.write(line)
for cv, pmf in np.nditer([cv_czar, pmf_total]):
fp_out.write(f'{cv:10.4f} {pmf:12.7f}\n')
#np.savetxt(output_pmf, np.c_[cv_czar, pmf_total], fmt = '%10.4f %12.7f')
if plot is True:
# reference PMF
ref_cv, ref_pmf = np.genfromtxt('../ref.dat', unpack=True)
# WTM-eABF PMF
wtm_cv, wtm_pmf = np.genfromtxt('wtm-eabf.pmf', unpack=True)
# plotting
w, h = figaspect(1 / 1.2)
plt.figure(figsize=(w, h))
plt.plot(ref_cv, ref_pmf, color='black', label='Reference')
plt.plot(wtm_cv, wtm_pmf, color='tab:orange', label='WTM-eABF')
plt.plot(cv_czar, pmf_total, color='tab:red', label='eABF + GaMDD')
plt.plot(cv_czar, pmf_czar, color='tab:blue', label='eABF part')
plt.plot(cv_czar, pmf_amd, color='tab:green', label='GaMDD part')
plt.xlabel('Distance (nm)')
plt.ylabel('$\Delta G$ (kcal/mol)')
ax = plt.gca()
ax.tick_params(direction='in',
which='major',
length=6.0,
width=1.0,
top=True,
right=True)
ax.tick_params(direction='in',
which='minor',
length=3.0,
width=1.0,
top=True,
right=True)
ax.xaxis.get_major_formatter()._usetex = False
ax.yaxis.get_major_formatter()._usetex = False
ax.xaxis.set_minor_locator(AutoMinorLocator())
ax.yaxis.set_minor_locator(AutoMinorLocator())
ax.set_ylim(0, 30)
plt.legend(prop={'size': 14}, fancybox=False, frameon=False)
plt.tight_layout(pad=0.2)
output_png = outputname + '.png'
plt.savefig(output_png, dpi=600, transparent=False)
plt.close()
def hist(scale):
x = np.random.normal(size=1000000, scale=scale)
fig = plt.figure(figsize=figaspect(1))
ax = fig.gca()
ax.hist(x, range=(-6, 6), bins=200, density=True)
ax.set(xlim=(-6, 6), ylim=(0, 1))
plt.show()
def __init__(self, error_a, error_v, angle, angle_no_small):
plt.close()
w, h = figaspect(1 / 2.9)
fix, axes = plt.subplots(1, 3, figsize=(w, h))
self.fontsize_title = 20
self.fontsize_label = 15
plt.tight_layout(pad=3.1, w_pad=3)
self.nbins = 32
self.n1 = self.plot_error_annotation(axes[0],
data=error_a,
title='Arousal')
self.n2 = self.plot_error_annotation(axes[1],
data=error_v,
title='Valence')
self.plot_error_angle(axes[2], data=angle, title='Emotion Angle')
# self.plot_error_angle(axes[3], data=angle_no_small, title='Emotion Angle No Small')
# axes[0].axvline(x=-0.225, color='red', linestyle='dashed')
# axes[0].axvline(x=0.225, color='red', linestyle='dashed')
# axes[1].axvline(x=-0.230, color='red', linestyle='dashed')
# axes[1].axvline(x=0.230, color='red', linestyle='dashed')
for ax in axes:
ax.set_ylim(axes[2].get_ylim())
ax.tick_params(axis='both', which='major', labelsize=15)
plt.savefig('../error_stats_angle.pdf', format='pdf')
plt.show()
def plot_dSm(data, image_index, outputname):
w, h = figaspect(1 / 2)
plt.figure(figsize=(w, h))
x = np.arange(0, len(data), 1)
plt.plot(x, data)
plt.title(f'm = {image_index+1}')
plt.xlabel(r'Iterations')
plt.ylabel(r'$\sqrt{S^m(\delta\tau)}/\delta\tau$ (°)')
ax = plt.gca()
ax.set_ylim(0, 1.5)
ax.tick_params(direction='in',
which='major',
length=6.0,
width=1.0,
top=True,
right=True)
ax.tick_params(direction='in',
which='minor',
length=3.0,
width=1.0,
top=True,
right=True)
ax.xaxis.set_minor_locator(AutoMinorLocator())
ax.yaxis.set_minor_locator(AutoMinorLocator())
ax.xaxis.get_major_formatter()._usetex = False
ax.yaxis.get_major_formatter()._usetex = False
plt.savefig(outputname, dpi=300, bbox_inches='tight', transparent=False)
plt.close()
def show():
x_min, x_max = -3, 3
coef = 2
sigma = 1
x = np.arange(x_min + 1, x_max)
np.random.seed(1234)
e = np.random.normal(scale=sigma, size=x.size)
fig = plt.figure(figsize=figaspect(1))
ax = fig.gca()
ax.plot((x_min, x_max), (x_min * coef, x_max * coef),
c='k',
label='$f(X_{i})$')
ax.scatter(x, x * coef + e, c='k', label='$(X_{i},\ y_{i})$')
for i, y_hat in enumerate(x * coef):
x_tmp = x_min + i + 1
ee = np.linspace(-sigma * 3, sigma * 3, 10)
xx = x_tmp + norm.pdf(ee)
yy = y_hat + ee
lines = ax.plot(xx, yy, c='b')
line_collection = ax.hlines(y_hat + e[i],
x_tmp,
x_tmp + norm.pdf(e[i]),
color='r')
if i is 0:
lines[0].set(label=r'$\epsilon _{i}$')
line_collection.set(label=r'$P(y_{i}\ |\ X_{i})$')
ax.legend()
ax.set(xlim=(x_min, x_max), xticks=(), yticks=())
plt.show()
def gen_steam_stats_graph(self, data):
graph_data = data['graph']
steps = timedelta(milliseconds=graph_data['step'])
timestamp = datetime.utcfromtimestamp(graph_data['start'] / 1000)
plots = graph_data['data']
times = []
for _ in plots:
timestamp -= steps
times.append(timestamp)
plt.style.use('dark_background')
w, h = figaspect(1 / 3)
fig, ax = plt.subplots(figsize=(w, h))
ax.grid(linestyle='-', linewidth='0.5', color='white')
plt.setp(plt.plot(list(reversed(times)), plots, linewidth=4),
color='#00adee')
plt.title(
f'Steam CM status over the last {human_timedelta(timestamp)[:-4]}',
size=20)
plt.axis([None, None, 0, 100])
plt.xlabel('Time (Month-Day Hour)', fontsize=20)
plt.ylabel('Uptime (%)', fontsize=20)
plt.tight_layout(h_pad=20, w_pad=20)
buf = BytesIO()
plt.savefig(buf, format='png', transparent=True)
buf.seek(0)
plt.close()
return discord.File(buf, filename='graph.png')
def show(x, y):
model = LinearSVC(C=1e30, random_state=1234)
model.fit(x, y)
fig = plt.figure(figsize=figaspect(1))
ax = fig.gca()
ax.scatter(x[:, 0], x[:, 1], c=y)
xlim = ax.get_xlim()
ylim = ax.get_ylim()
xx, yy = np.meshgrid(np.linspace(xlim[0], xlim[1], resolution),
np.linspace(ylim[0], ylim[1], resolution))
grid = np.c_[xx.ravel(), yy.ravel()]
zz = model.decision_function(grid).reshape(xx.shape)
cs = ax.contour(xx,
yy,
zz,
colors='k',
levels=[-1, 0, 1],
linewidths=0.8,
linestyles=['--', '-', '--'])
fmt = {
c: '決定境界' if i is 1 else 'サポートベクター'
for i, c in enumerate(cs.levels)
}
ax.clabel(cs, fontsize='large', fmt=fmt)
ax.set(xlim=xlim, ylim=ylim, xticks=(), yticks=())
plt.show()
def plotNumpy3D(self, arr, color=None, time=None, size=None):
plt.clf()
w, h = figaspect(1)
w *= 1.6
h = w
fig = plt.figure(figsize=(w, h))
ax = fig.add_subplot(111, projection='3d')
ax.view_init(azim=100, elev=60)
# ax = fig.add_subplot(111) # not 3D
ax.grid(linestyle='--')
ax.set_axisbelow(True)
ax.tick_params(axis='both',
which='both',
left=False,
top=False,
right=False,
bottom=False)
x = arr[:, 0]
y = arr[:, 1]
z = arr[:, 2]
ax.scatter(x,
y,
z,
c=('C0' if color is None else color),
s=(10 if size is None else size))
ax.scatter(self.gnList[:, 0],
self.gnList[:, 1],
self.gnList[:, 2],
c='green',
s=50,
alpha=0.3)
# ax.scatter(x,y,c=('C0' if color is None else color), s=100)# not 3D
# ax.scatter(self.gnList[:,0],self.gnList[:,1],c='green', s=50, alpha=0.2)# not 3D
for i in arr:
realRadiusRange = ((((self.longRange)**2) + (z[0]**2))**0.5)
# p = Circle((i[0],i[1]), self.longRange, color=color, alpha=0.3, lw=0.8)
p = Circle((i[0], i[1]),
realRadiusRange,
color=color,
alpha=0.3,
lw=0.8)
ax.add_patch(p)
art3d.pathpatch_2d_to_3d(p, z=1.5, zdir="z")
# ax.set_xlim(-10,120)
# ax.set_ylim(-10,120)
if time is not None:
if time > 0:
plt.pause(time)
plt.clf()
else:
plt.show()
def show():
bins = [10, 30]
tip = sns.load_dataset('tips')['tip']
_, axes = plt.subplots(1, 2, figsize=figaspect(1 / 2))
for ax, b in zip(axes, bins):
ax.hist(tip, bins=b)
ax.set(title=f'bins={b}', xticks=(), yticks=())
plt.show()
def plot(df):
chi2 = stats.chi2.pdf(x, df=df)
plt.figure(figsize=figaspect(1))
ax = plt.axes()
ax.plot(x, chi2, label=r'自由度 ${df}$ の $\chi^2$ 分布'.format(df=df))
ax.legend()
ax.set(title=r'$\chi^2$ 分布', xlim=xlim, ylim=(0, 0.3))
plt.show()
def _process(self, img, savepath=None, **kwargs): if savepath is not None: H, W, _ = self.base.shape w, h = figaspect(H / W) w, h = self.img_scale * w, self.img_scale * h plt.savefig(savepath, dpi=W / w) if not self._view: plt.close(plt.gcf())
def refreshGraph(graph, node_color, fig):
print("here")
#plt.clf()
pos = nx.get_node_attributes(graph, 'pos')
w, h = figaspect(5 / 3)
fig, ax = plt.subplots(figsize=(w, h))
nx.draw(graph, pos, node_color=node_color, node_size=20, ax=ax)
fig.canvas.draw()
def plot_1d(data, outputname, xlabel, ylabel):
plt.rcParams.update({
"pgf.texsystem":
"lualatex",
"font.family":
"serif", # use serif/main font for text elements
"text.usetex":
True, # use inline math for ticks
"pgf.rcfonts":
False, # don't setup fonts from rc parameters
"axes.labelsize":
28,
"axes.linewidth":
2.0,
'axes.unicode_minus':
False,
"font.size":
24,
"pgf.preamble":
'\n'.join([
"\\usepackage{units}",
"\\usepackage{metalogo}",
"\\usepackage{unicode-math}",
r"\setmathfont{MathJax_Math}",
r"\setmainfont{FreeSans}",
])
})
w, h = figaspect(1 / 1.1)
plt.figure(figsize=(w, h))
plt.plot(data[0], data[1])
# plt.title(title)
plt.xlabel(xlabel)
plt.ylabel(ylabel)
ax = plt.gca()
ax.set_xlim(-180.0, 180.0)
ax.set_ylim(0, 16.0)
ax.xaxis.get_major_formatter()._usetex = False
ax.yaxis.get_major_formatter()._usetex = False
ax.tick_params(direction='in',
which='major',
length=6.0,
width=1.0,
top=True,
right=True,
pad=8.0)
ax.tick_params(direction='in',
which='minor',
length=3.0,
width=1.0,
top=True,
right=True,
pad=8.0)
ax.xaxis.set_major_locator(plt.MultipleLocator(90))
ax.yaxis.set_major_locator(AutoLocator())
ax.xaxis.set_minor_locator(AutoMinorLocator())
ax.yaxis.set_minor_locator(AutoMinorLocator())
plt.savefig(outputname, dpi=300, bbox_inches='tight', transparent=False)
plt.close()
def plot(i):
_, (ax1, ax2) = plt.subplots(1,
2,
figsize=figaspect(1 / 2),
sharex=True)
plot_cdf(ax1, i, norm.cdf(i))
ax1.set(xticklabels=['in'], yticklabels=[0, 'out', 1])
plot_pdf_area(ax2, np.linspace(x_min, i, resolution))
plt.show()
def plotEtaZeta(t_list, eta_list, zeta_list):
# plot AC rate
w, h = figure.figaspect(6.)
fig, axes = plt.subplots(figsize=(h, w))
axes.set_xlabel('hours')
axes.plot(t_list, zeta_list, color='steelblue', lw=0.7)
axes.set_ylabel('AC rate [mas/s]')
plt.savefig('zetaVStime.png')
plt.close()
# plot AL rate
w, h = figure.figaspect(6.)
fig, axes = plt.subplots(figsize=(h, w))
axes.set_xlabel('hours')
axes.plot(t_list, eta_list, color='steelblue', lw=0.7)
axes.set_ylabel('AL rate [mas/s]')
plt.savefig('etaVStime.png')
plt.close()
def plot(count):
distributions = stats.norm(loc=sample[:count])
fig = plt.figure(figsize=figaspect(1))
ax = fig.gca()
ax.plot(x, distributions.pdf(x))
ax.plot(x, distributions.pdf(x).sum(axis=1), color='black', alpha=0.3)
sns.rugplot(sample, color='black', ax=ax)
ax.set(xlabel='x', xticks=(), xlim=(x.min(), x.max()), ylim=(0, 1))
plt.show()
def plot(df):
t = stats.t.pdf(x, df=df)
plt.figure(figsize=figaspect(1))
ax = plt.axes()
ax.plot(x, n, label='標準正規分布')
ax.plot(x, t, label=f'自由度 ${df}$ の $t$ 分布')
ax.legend()
ax.set(title='標準正規分布と $t$ 分布', xlim=xlim, ylim=ylim)
plt.show()
def plot(i):
_, (ax1, ax2) = plt.subplots(1,
2,
figsize=figaspect(1 / 2),
sharex=True)
plot_cdf(ax1, i, 1 - norm.sf(i), y_origin=1)
ax1.set(xticklabels=['in'], yticklabels=[0, '', 1])
plot_pdf_area(ax2, np.linspace(i, x_max, resolution))
plt.show()
def plot(total, n_objects, n_drawn):
x = np.arange(0, n_drawn + 1)
distribution = stats.hypergeom(total, n_objects, n_drawn)
plt.figure(figsize=figaspect(1))
ax = plt.axes()
ax.vlines(x, 0, distribution.pmf(x))
ax.set(title='超幾何分布', xlabel='対象の個数', ylim=(0, 1), xticks=x)
ax.set_ylabel('確率', rotation=0, horizontalalignment='right')
plt.show()
def make_yyplot( ActualY, EstimatedY, YMax, YMin, EstimatedYName ):
plt.figure(figsize=figure.figaspect(1))
plt.scatter(ActualY,EstimatedY)
plt.plot([YMin-0.05*(YMax-YMin),YMax+0.05*(YMax-YMin)], [YMin-0.05*(YMax-YMin),YMax+0.05*(YMax-YMin)], 'k-')
plt.ylim(YMin-0.05*(YMax-YMin),YMax+0.05*(YMax-YMin))
plt.xlim(YMin-0.05*(YMax-YMin),YMax+0.05*(YMax-YMin))
plt.xlabel("Actual Y")
plt.ylabel(EstimatedYName)
plt.show()
def plot(int_range):
x_min, x_max = int_range
x = np.arange(x_min, x_max + 1)
plt.figure(figsize=figaspect(1))
ax = plt.axes()
ax.vlines(x, 0, 1.0 / x.size)
ax.set(title='離散一様分布', xlabel='取りうる値', xlim=xlim, ylim=(0, 1.1), xticks=x)
ax.set_ylabel('確率', rotation=0, horizontalalignment='right')
plt.show()
def plot(p):
x = [0, 1]
y = stats.bernoulli.pmf(x, p)
plt.figure(figsize=figaspect(1))
ax = plt.axes()
ax.vlines(x, 0, y)
ax.set(title='ベルヌーイ分布', xlabel='取りうる値', ylim=(0, 1.1), xticks=x)
ax.set_ylabel('確率', rotation=0, horizontalalignment='right')
plt.show()
def figure_setup(num_subplots, aspect, scale=1):
#plt.style.use('pluto-paper')
fig, axs = plt.subplots(
nrows=num_subplots,
sharex=True,
gridspec_kw={'hspace': 0},
figsize=[scale * dim for dim in figaspect(num_subplots * aspect)])
if num_subplots == 1:
axs = [axs]
return fig, axs
def convert_data_to_graph(self, visualize=True):
# Create network graph
G = nx.Graph()
list_of_nodes = []
current_id = 0
for entity in self.line_data:
connectivity_list = []
for point in entity:
point = tuple([int(_) for _ in point])
# If we are beginning then list is empty (i.e. no nodes yet)
if not list_of_nodes:
list_of_nodes.append(Node(current_id, point))
connectivity_list.append(current_id)
current_id += 1
else:
bFoundSameNode = False
for node in list_of_nodes:
# Check if the point in question is the same as a node
if point == node.coordinate:
connectivity_list.append(node.id)
bFoundSameNode = True
break
if not bFoundSameNode:
list_of_nodes.append(Node(current_id, point))
connectivity_list.append(current_id)
current_id += 1
for con_idx, connection in enumerate(connectivity_list[:-1]):
for node in list_of_nodes:
if node.id == connection:
node.connected_to.append(connectivity_list[con_idx +
1])
for node in list_of_nodes:
G.add_node(node.id, pos=node.coordinate)
for node in list_of_nodes:
if not node.connected_to:
pass
else:
for node_connection in node.connected_to:
'''
p0 = np.asarray(list_of_nodes[node.id].coordinate)
p1 = np.asarray(list_of_nodes[node_connection].coordinate)
weight = np.linalg.norm(p0 - p1)
G.add_edge(node.id, node_connection, weight=weight)
'''
G.add_edge(node.id, node_connection)
if visualize:
pos = nx.get_node_attributes(G, 'pos')
w, h = figaspect(5 / 3)
fig, ax = plt.subplots(figsize=(w, h))
nx.draw(G, pos, node_size=20, ax=ax)
plt.show()
return G
def costmap_plot(arr):
# using costmap_plot() to plot the background first (with resolution of (map_size/map_res)^2), then in update_plot the color(zz, as probability) is updated.
w, h = figaspect(.8)
fig = plt.figure(figsize=(w,h))
ani = animation.FuncAnimation(fig, update_plot, interval = 1)
plt.show()
def __init__(self, grid, animation_step, step_interval, arrow_for_potential):
self.grid = grid
self.animation_step = animation_step
self.arrow_potential = arrow_for_potential
#Setup subplots
aspect = figaspect(1)
self.figure, self.axis = plt.subplots(figsize=1.5 * aspect)
self.arrows = self.place_arrows()
self.animated_matrix = self.axis.matshow(self.grid.mat, vmin=Cell.RESERVED, vmax=Direction.RIGHT,
cmap=cm.Paired)
self.animation = animation.FuncAnimation(self.figure, self.update_view, self.animation_step,
init_func=self.update_view,
interval=step_interval,
save_count=0, blit=True)
def view(structure, outdoor_lux=109870., indoor_lux=500., show_original=False):
'''
Experimental. Lux ratio is is buggy
Default outdoor lux is for AM1.5 and indoor lux for office lighting.
See https://en.wikipedia.org/wiki/Daylight and https://en.wikipedia.org/wiki/Lux
'''
if not structure._color_calculated:
structure.calculate_color()
T_image = mpimg.imread(sup_path + "test_outdoor.png")
R_image = mpimg.imread(sup_path + "test_indoor.png")
T_filter = np.array(structure.T_color, float)
R_filter = np.array(structure.R_color, float)
T_image_after = T_image * T_filter
R_image_after = R_image * R_filter
if outdoor_lux > indoor_lux:
R_image_after *= indoor_lux / outdoor_lux
else:
T_image_after *= outdoor_lux / indoor_lux
overlay = T_image_after + R_image_after
f, ax = plt.subplots(nrows=1, ncols=1)
f.suptitle(structure.label)
ax.set_title("Outdoor lux: {:d} Indoor lux: {:d}".format(int(outdoor_lux), int(indoor_lux)))
ax.imshow(overlay)
ax.axis("off")
if show_original:
w, h = figaspect(2.)
f2, ((ax1), (ax2)) = plt.subplots(nrows=2, ncols=1, sharex='col', figsize=(w,h))
f2.subplots_adjust(left=0.03, right=0.97, hspace=0.0, wspace=0.0)
ax1.axis("off")
ax2.axis("off")
ax1.set_aspect("equal")
ax2.set_aspect("equal")
ax1.imshow(T_image)
ax2.imshow(R_image)
Y_pred_ceiling = sp.maximum(Y_pred_ceiling, 1)
kappa_cdf_cv[run, fold] = quadratic_weighted_kappa(Y_valid, Y_pred_cdf)
kappa_rounding_cv[run, fold] = quadratic_weighted_kappa(Y_valid, Y_pred_rounding)
kappa_ceiling_cv[run, fold] = quadratic_weighted_kappa(Y_valid, Y_pred_ceiling)
print("Kappa using cdf decoding method: %.6f (%.6f)" % (np.mean(kappa_cdf_cv), np.std(kappa_cdf_cv)))
print("Kappa using rounding decoding method: %.6f (%.6f)" % (np.mean(kappa_rounding_cv), np.std(kappa_rounding_cv)))
print("Kappa using ceiling decoding method: %.6f (%.6f)" % (np.mean(kappa_ceiling_cv), np.std(kappa_ceiling_cv)))
###################
## Visualization ##
###################
w, h = figaspect(1)
plt.figure(figsize=(w,h))
f, axarr = plt.subplots(4,4, sharex=True)
## raw prediction
axarr[0,0].hist(Y_pred_raw[Y_valid==1])
axarr[0,0].set_title("Relevance = 1")
axarr[0,0].set_ylabel("Raw")
axarr[0,1].hist(Y_pred_raw[Y_valid==2])
axarr[0,1].set_title("Relevance = 2")
axarr[0,2].hist(Y_pred_raw[Y_valid==3])
axarr[0,2].set_title("Relevance = 3")
axarr[0,3].hist(Y_pred_raw[Y_valid==4])
axarr[0,3].set_title("Relevance = 4")
## rounding decoding
axarr[1,0].hist(Y_pred_rounding[Y_valid==1])
axarr[1,0].set_ylabel("Rounding")
def main():
# collect argvs
log_file = sys.argv[1]
if len(sys.argv) > 2:
pdf_file = sys.argv[2]
else:
pdf_file = sys.argv[1][:-4] + ".pdf"
# fetch training and validation iteration
cmd = (
"cat %s | grep 'solver.cpp:231] Iteration ' | awk '{print $6}' | awk -F',' '{print $1}' > train_iteration.tmp"
% log_file
)
subprocess.call(cmd, shell=True)
cmd = (
"cat %s | grep 'solver.cpp:287] Iteration ' | awk '{print $6}' | awk -F',' '{print $1}' > valid_iteration.tmp"
% log_file
)
subprocess.call(cmd, shell=True)
# fetch training and validation logloss
cmd = "cat %s | grep 'Train net output #3: loss_fine = ' | awk '{print $11}' > train_logloss.tmp" % log_file
subprocess.call(cmd, shell=True)
cmd = "cat %s | grep 'Test net output #3: loss_fine =' | awk '{print $11}' > valid_logloss.tmp" % log_file
subprocess.call(cmd, shell=True)
# fetch training and validation accuracy
cmd = "cat %s | grep 'Train net output #1: accuracy_fine = ' | awk '{print $11}' > train_accuracy.tmp" % log_file
subprocess.call(cmd, shell=True)
cmd = "cat %s | grep 'Test net output #1: accuracy_fine =' | awk '{print $11}' > valid_accuracy.tmp" % log_file
subprocess.call(cmd, shell=True)
train_logloss = np.loadtxt("train_logloss.tmp")
valid_logloss = np.loadtxt("valid_logloss.tmp")
train_accuracy = 100 * np.loadtxt("train_accuracy.tmp")
valid_accuracy = 100 * np.loadtxt("valid_accuracy.tmp")
train_iteration = np.loadtxt("train_iteration.tmp", dtype=int)
valid_iteration = np.loadtxt("valid_iteration.tmp", dtype=int)
valid_logloss_min = np.min(valid_logloss)
valid_accuracy_min = np.min(valid_accuracy)
best_ind = np.where(valid_logloss == valid_logloss_min)[0][0]
best_iter = valid_iteration[best_ind]
# plot training and validation logloss
w, h = figaspect(0.5)
plt.figure(figsize=(w, h))
f, axarr = plt.subplots(2, sharex=True)
# logloss
axarr[0].plot(train_iteration, train_logloss)
axarr[0].plot(valid_iteration, valid_logloss)
axarr[0].plot(train_iteration, valid_logloss_min * np.ones((len(train_iteration))))
axarr[0].set_title("LogLoss vs Iteration")
axarr[0].set_ylabel("LogLoss")
axarr[0].legend(
["Train", "Valid (Min = %s at Iter. = %s)" % (np.round(valid_logloss_min, 5), best_iter)], loc="upper right"
)
# accuracy
axarr[1].plot(train_iteration, train_accuracy)
axarr[1].plot(valid_iteration, valid_accuracy)
axarr[1].plot(train_iteration, valid_accuracy[best_ind] * np.ones((len(train_iteration))))
axarr[1].set_title("Accuracy vs Iteration")
axarr[1].set_xlabel("Iteration")
axarr[1].set_ylabel("Accuracy [%]")
axarr[1].legend(
["Train", "Valid (Acc = %s%% at Iter. = %s)" % (np.round(valid_accuracy[best_ind], 2), best_iter)],
loc="lower right",
)
# fmt = '%.0f%%'
# yticks = mtick.FormatStrFormatter(fmt)
# axarr[1].yaxis.set_major_formatter(yticks)
# plt.tight_layout()
plt.savefig(pdf_file)
print("Save pdf figure to %s" % pdf_file)
# cleanup
cmd = "rm -rf train_iteration.tmp"
subprocess.call(cmd, shell=True)
cmd = "rm -rf valid_iteration.tmp"
subprocess.call(cmd, shell=True)
cmd = "rm -rf train_logloss.tmp"
subprocess.call(cmd, shell=True)
cmd = "rm -rf valid_logloss.tmp"
subprocess.call(cmd, shell=True)
cmd = "rm -rf train_accuracy.tmp"
subprocess.call(cmd, shell=True)
cmd = "rm -rf valid_accuracy.tmp"
subprocess.call(cmd, shell=True)
def runtest(lmaManager=None, lma_view=None, HDFmanagers=None):
# colormap = get_cmap('gist_yarg_r')
colormap = get_cmap('gist_earth')
density_maxes = []
total_counts = []
all_t = []
for delta_minutes in minute_intervals:
time_delta = DateTimeDelta(0, 0, delta_minutes, 0)
n_frames = int(ceil((end_time - start_time) / time_delta))
n_cols = 6
n_rows = int(ceil( float(n_frames) / n_cols ))
w, h = figaspect(float(n_rows)/n_cols)
xedge=np.arange(b.x[0], b.x[1]+dx, dx)
yedge=np.arange(b.y[0], b.y[1]+dy, dy)
x_range = b.x[1] - b.x[0]
y_range = b.y[1] - b.y[0]
min_count, max_count = 1, max_count_baseline*delta_minutes
f = figure(figsize=(w,h))
p = small_multiples_plot(fig=f, rows=n_rows, columns=n_cols)
p.label_edges(True)
for ax in p.multiples.flat:
ax.yaxis.set_major_formatter(kilo_formatter)
ax.xaxis.set_major_formatter(kilo_formatter)
for i in range(n_frames):
frame_start = start_time + i*time_delta
frame_end = frame_start + time_delta
b.sec_of_day = (frame_start.abstime, frame_end.abstime)
b.t = (frame_start, frame_end)
do_plot = False
flash_extent_density = True
density = None
if source_density==True:
lmaManager.refresh(b)
lma_view.transformed.cache_is_old()
x,y,t=lma_view.transformed['x','y','t']
density,edges = np.histogramdd((x,y), bins=(xedge,yedge))
do_plot=True
else:
for lmaManager in HDFmanagers:
# yes, loop through every file every time and reselect data.
# so wrong, yet so convenient.
h5 = lmaManager.h5file
if flash_extent_density == False:
lmaManager.refresh(b)
lma_view = AcuityView(DataSelection(lmaManager.data, b), mapProj, bounds=b)
# lma_view.transformed.cache_is_old()
x,y,t=lma_view.transformed['x','y','t']
if x.shape[0] > 1: do_plot = True
break
else:
# assume here that the bounds sec_of_day day is the same as
# the dataset day
t0, t1 = b.sec_of_day
# events = getattr(h5.root.events, lmaManager.table.name)[:]
# flashes = getattr(h5.root.flashes, lmaManager.table.name)[:]
event_dtype = getattr(h5.root.events, lmaManager.table.name)[0].dtype
events_all = getattr(h5.root.events, lmaManager.table.name)[:]
flashes = getattr(h5.root.flashes, lmaManager.table.name)
def event_yielder(evs, fls):
these_events = []
for fl in fls:
if ( (fl['n_points']>9) &
(t0 < fl['start']) &
(fl['start'] <= t1)
):
these_events = evs[evs['flash_id'] == fl['flash_id']]
if len(these_events) <> fl['n_points']:
print 'not giving all ', fl['n_points'], ' events? ', these_events.shape
for an_ev in these_events:
yield an_ev
# events = np.fromiter((an_ev for an_ev in ( events_all[events_all['flash_id'] == fl['flash_id']]
# for fl in flashes if (
# (fl['n_points']>9) & (t0 < fl['start']) & (fl['start'] <= t1)
# )
# ) ), dtype=event_dtype)
events = np.fromiter(event_yielder(events_all, flashes), dtype=event_dtype)
# print events['flash_id'].shape
### Flash extent density ###
x,y,z = mapProj.fromECEF(
*geoProj.toECEF(events['lon'], events['lat'], events['alt'])
)
# Convert to integer grid coordinate bins
# 0 1 2 3
# | | | | |
# -1.5 0.0 1.5 3.0 4.5
if x.shape[0] > 1:
density, edges = extent_density(x,y,events['flash_id'].astype('int32'),
b.x[0], b.y[0], dx, dy, xedge, yedge)
do_plot = True
break
# print 'density values: ', density.min(), density.max()
if do_plot == True: # need some data
# density,edges = np.histogramdd((x,y), bins=(xedge,yedge))
density_plot = p.multiples.flat[i].pcolormesh(xedge,yedge,
np.log10(density.transpose()),
vmin=-0.2,
vmax=np.log10(max_count),
cmap=colormap)
label_string = frame_start.strftime('%H%M:%S')
text_label = p.multiples.flat[i].text(b.x[0]-pad+x_range*.01, b.y[0]-pad+y_range*.01, label_string, color=(0.5,)*3, size=6)
density_plot.set_rasterized(True)
density_maxes.append(density.max())
total_counts.append(density.sum())
all_t.append(frame_start)
print label_string, x.shape, density.max(), density.sum()
color_scale = ColorbarBase(p.colorbar_ax, cmap=density_plot.cmap,
norm=density_plot.norm,
orientation='horizontal')
# color_scale.set_label('count per pixel')
color_scale.set_label('log10(count per pixel)')
# moving reference frame correction. all panels will have same limits, based on time of last frame
view_dt = 0.0 # (frame_start - t0).seconds
x_ctr = x0 + view_dt*u
y_ctr = y0 + view_dt*v
view_x = (x_ctr - view_dx/2.0 - pad, x_ctr + view_dx/2.0 + pad)
view_y = (y_ctr - view_dy/2.0 - pad, y_ctr + view_dy/2.0 + pad)
# view_x = (b.x[0]+view_dt*u, b.x[1]+view_dt*u)
# view_y = (b.y[0]+view_dt*v, b.y[1]+view_dt*v)
# print 'making timeseries',
# time_series = figure(figsize=(16,9))
# ts_ax = time_series.add_subplot(111)
# ts_ax.plot_date(mx2num(all_t),total_counts,'-', label='total sources', tz=tz)
# ts_ax.plot_date(mx2num(all_t),density_maxes,'-', label='max pixel', tz=tz)
# ts_ax.xaxis.set_major_formatter(time_series_x_fmt)
# ts_ax.legend()
# time_filename = 'out/LMA-timeseries_%s_%5.2fkm_%5.1fs.pdf' % (start_time.strftime('%Y%m%d_%H%M%S'), dx/1000.0, time_delta.seconds)
# time_series.savefig(time_filename)
# print ' ... done'
print 'making multiples',
p.multiples.flat[0].axis(view_x+view_y)
filename = 'out/LMA-density_%s_%5.2fkm_%5.1fs.pdf' % (start_time.strftime('%Y%m%d_%H%M%S'), dx/1000.0, time_delta.seconds)
f.savefig(filename, dpi=150)
print ' ... done'
f.clf()
return events