# Construct and compile neural event model
ne_clf = manifold.TSNE(n_components=2)
ne_embedded = nc_clf.fit_transform(train_proc_neural)
# Plot both embeddings in the same figure
f, axs = plt.subplots(3)
for idx, bc in enumerate(np.unique(behavior_clusters)):
axs[0].scatter(be_embedded[behavior_clusters == bc, 0],
be_embedded[behavior_clusters == bc, 1],
c=cm.Reds_r(idx))
axs[0].set_title('Behavior embeddings colored by state.')
for idx, nc in enumerate(np.unique(neural_clusters)):
axs[1].scatter(ne_embedded[neural_clusters == nc, 0],
ne_embedded[neural_clusters == nc, 1],
c=cm.Greens_r(idx))
axs[1].set_title('Neural embeddings colored by state.')
axs[2].plot(be_embedded[:, 0], be_embedded[:, 1], 'r', alpha=1)
axs[2].plot(ne_embedded[:, 0], ne_embedded[:, 1], 'g', alpha=0.6)
# Write a brief report and save associated data.
report = 'We applied a cluster analysis to behavior and neural ' +\
'data, identifying "states" in each. There are different ' +\
'numbers of clusters in each. This suggests that better ' +\
'decoding performance could be achieved by selecting ' +\
'the behavior and neural dimensions that best harmonize ' +\
'the two data sources.'
# Make predictions on the test set and save for the competition
fn = c_utils.savefig(team_name)
c_utils.movefile_for_eval(fn)
frame = event.frame
image_feat = recog_net.feat_extract(frame)
image_feature[counter, :] = image_feat.data.squeeze().unsqueeze_(0).numpy()
counter += 1
event = env.step(action=dict(action='RotateRight', rotation=90.0))
# low dimensional embedding
svd = TruncatedSVD(n_components=50, n_iter=10)
image_feature_SVD = svd.fit_transform(image_feature)
image_embedded = TSNE(n_components=2).fit_transform(image_feature_SVD)
# plot embeddings
ff1 = plt.figure(figsize=(8, 6))
counter = 0
for i in range(100):
plt.scatter(image_embedded[counter,0], image_embedded[counter,1],c=cm.Purples_r(i / 100.0))
counter += 1
for i in range(100):
plt.scatter(image_embedded[counter,0], image_embedded[counter,1],c=cm.Greens_r(i / 100.0))
counter += 1
for i in range(100):
plt.scatter(image_embedded[counter,0], image_embedded[counter,1],c=cm.Blues_r(i / 100.0))
counter += 1
for i in range(100):
plt.scatter(image_embedded[counter,0], image_embedded[counter,1],c=cm.Oranges_r(i / 100.0))
counter += 1
plt.tight_layout()
plt.savefig('embedding3_ResNet3.png')
plt.close(ff1)
def plot(self):
import matplotlib.pyplot as plt
from matplotlib import cm
import matplotlib.dates as mdates
import seaborn as sns
sns.set_style("darkgrid")
start_date = "2007-01-01" # Day 1 of simulation
date_format = "%Y-%m-%d"
foo = mdates.strpdate2num(date_format)
daynum = np.arange(self.n_tstep)
daydates_list = []
daydates_mdates = np.array([])
for dayn in daynum:
hold = convert_to_date(dayn, start_date, date_format=date_format)
daydates_list.append(hold)
daydates_mdates = np.append(daydates_mdates, foo(hold))
print(daydates_mdates)
maxpop_ever = 0
for sim_id, data in self.RDT_prev_by_node.items():
for j in range(self.n_nodes):
maxp = np.max(self.pop_by_node[sim_id][j])
if maxp > maxpop_ever:
maxpop_ever = maxp
plt.figure(figsize=(12, 5))
for sim_id, data in self.RDT_prev_by_node.items():
for j in range(self.n_nodes):
maxp = np.max(self.pop_by_node[sim_id][j])
pop_ratio = np.float(maxp) / np.float(maxpop_ever)
plt.plot_date(daydates_mdates,
self.RDT_prev_by_node[sim_id][j],
fmt='-',
alpha=pop_ratio,
color=cm.Greens_r(pop_ratio),
zorder=1)
catch = self.catch.itervalues().next()
# Look up catchment prevalence data from precomputed file:
base = 'C:/Users/jsuresh/OneDrive - IDMOD/Projects/zambia-gridded-sims/'
df = pd.read_csv(
base +
"data/interventions/kariba/2017-11-27/cleaned/catch_prevalence.csv"
)
catch_prev = np.array(df[catch])
round_dates = [
"2012-07-01", "2012-09-30", "2012-11-30", "2013-07-01",
"2013-08-31", "2013-10-31", "2014-12-31", "2015-03-01",
"2015-09-30", "2016-02-29"
]
# round_dates = {
# "1": "2012-07-01",
# "2": "2012-09-30",
# "3": "2012-11-30",
# "4": "2013-07-01",
# "5": "2013-08-31",
# "6": "2013-10-31",
# "7": "2014-12-31",
# "8": "2015-03-01",
# "9": "2015-09-30",
# "10":"2016-02-29"
# }
round_dates_mdate = []
for i in range(10):
day_mdate = foo(round_dates[i])
round_dates_mdate.append(day_mdate)
round_dates_array = np.array(round_dates_mdate)
if catch in [
"chabbobboma", "chipepo", "gwembe", "lukande", "nyanga chaamwe"
]:
plt.scatter(round_dates_array[:-4],
catch_prev[:-4],
c='red',
s=70,
label='Data for {}'.format(catch.capitalize()),
zorder=10)
plt.scatter(round_dates_array[-4:],
catch_prev[-4:],
c='gray',
s=70,
label='HFCA not in MDA round',
zorder=10)
elif catch in ["chisanga"]:
plt.scatter(np.append(round_dates_array[:3],
round_dates_array[5:]),
np.append(catch_prev[:3], catch_prev[5:]),
c='red',
s=70,
label='Data for {}'.format(catch.capitalize()),
zorder=10)
plt.scatter(round_dates_array[3:5],
catch_prev[3:5],
c='gray',
s=70,
label='HFCA not in MDA round',
zorder=10)
else:
plt.scatter(round_dates_array,
catch_prev,
c='red',
s=70,
label='Data for {}'.format(catch.capitalize()),
zorder=10)
# # Plot Chiyabi interventions as vertical lines:
# if True:
# # plot_only_first = False
#
# # Event information files
# itn_event_file = "C:/Users/jsuresh/OneDrive - IDMOD/Code/zambia/cbever/chiyabi_hfca_itn_events.csv"
# irs_event_file = "C:/Users/jsuresh/OneDrive - IDMOD/Code/zambia/cbever/chiyabi_hfca_irs_events.csv"
# msat_event_file = "C:/Users/jsuresh/OneDrive - IDMOD/Code/zambia/cbever/chiyabi_hfca_msat_events.csv"
# mda_event_file = "C:/Users/jsuresh/OneDrive - IDMOD/Code/zambia/cbever/chiyabi_hfca_mda_events.csv"
# healthseek_event_file = "C:/Users/jsuresh//OneDrive - IDMOD/Code/zambia/cbever/chiyabi_hfca_healthseek_events.csv"
# stepd_event_file = "C:/Users/jsuresh//OneDrive - IDMOD/Code/zambia/cbever/chiyabi_hfca_stepd_events.csv"
#
# # Import event info
# itn_events = pd.read_csv(itn_event_file)
# irs_events = pd.read_csv(irs_event_file)
# msat_events = pd.read_csv(msat_event_file)
# mda_events = pd.read_csv(mda_event_file)
# healthseek_events = pd.read_csv(healthseek_event_file)
# stepd_events = pd.read_csv(stepd_event_file)
#
# # for date in itn_events['fulldate']: plt.axvline(convert_to_day(date, start_date, date_format=date_format), ls='dashed', color='C0') #,label='ITN Events')
# # for date in irs_events['fulldate']: plt.axvline(convert_to_day(date, start_date, date_format=date_format), ls='dashed', color='C1') #,label='IRS Events')
# # for date in msat_events['fulldate']: plt.axvline(convert_to_day(date, start_date, date_format=date_format), ls='dashed', color='C2') #,label='MSAT Events')
# # for date in mda_events['fulldate']: plt.axvline(convert_to_day(date, start_date, date_format=date_format), ls='dashed', color='C3') #,label='MDA Events')
#
# for date in itn_events['fulldate']: plt.axvline(foo(date), ls='dashed', color='blue') #,label='ITN Events')
# for date in irs_events['fulldate']: plt.axvline(foo(date), ls='dashed', color='green') #,label='IRS Events')
# for date in msat_events['fulldate']:plt.axvline(foo(date), ls='dashed', color='red') #,label='MSAT Events')
# for date in mda_events['fulldate']: plt.axvline(foo(date), ls='dashed', color='purple') #,label='MDA Events')
#
#
# # colors:
# # c0 = blue = ITN
# # c1 = green = IRS
# # c2 = red = MSAT
# # c3 = purple = MDA
plt.legend()
# plt.xlim([3000,7000])
plt.xlim([foo("2010-01-01"), foo("2019-01-01")])
# plt.show()
plt.tight_layout()
plt.savefig(base + "data/figs/{}_prev_node.png".format(catch))
img_all = np.concatenate(spl_list)
h2_all = tsne.fit_transform(np.concatenate(h2_list, axis=0))
hmin, hmax = h2_all.min(), h2_all.max()
width = np.ceil(hmax - hmin).astype(int)
wall = np.zeros([64 * width, 64 * width, 3])
for s in range(len(samples)):
x, y = np.int_(h2_all[s] - hmin) * 64
wall[x:x + 64, y:y + 64] = (samples[s] + 1) * 255 / 2
wall[wall < 0] = 0
wall[wall > 255] = 255
from matplotlib import cm
cstart = cm.Greens_r(np.linspace(0, 1, 255))
cstop = cm.Reds_r(np.linspace(0, 1, 255))
n_iter = 5000
cmaps = np.zeros([int(n_iter / 10), 255, 3])
for i in range(255):
for j in range(3):
cmaps[:, i, j] = np.linspace(cstart[i, j], cstop[i, j],
int(n_iter / 10))
h2_list, spl_list = [], []
for i_class in range(num_classes):
ppgn.sampler_init(i_class)
samples, h2 = ppgn.sample(i_class,
nbSamples=5000,
h2_start=None,
def plot_elasticity(self, spline, strain):
''' Plot the youngs modulus '''
out = np.array(self.youngs_modulus)
x = out[:, 0]
y = out[:, 1]
plt.figure()
plt.plot(x, y, 'x')
plt.title("Youngs Modulus")
plt.show()
naturalShape = strain.naturalShape
wholeShape = spline.spline
wholeShape_UpperHalf = []
for i in range(1, len(wholeShape[0]) - 1):
if (wholeShape[1][i] > 0.0):
wholeShape_UpperHalf.append(
[wholeShape[0][i], wholeShape[1][i]])
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
theta_grid = np.linspace(0, 2 * np.pi, 100)
out = np.array(naturalShape)
shape_x = out[:, 0]
shape_y = out[:, 1]
out = np.array(wholeShape_UpperHalf)
wholeShape_x = out[:, 0]
wholeShape_y = out[:, 1]
r_points, theta_points = np.meshgrid(shape_y, theta_grid)
r_wholeShape, theta_wholeShape = np.meshgrid(wholeShape_y, theta_grid)
temp = [self.youngs_modulus[-1][0], self.youngs_modulus[-1][1]]
self.youngs_modulus.append(temp)
out = np.array(self.youngs_modulus)
E_x = out[:, 0]
E_y = out[:, 1]
data, theta_points = np.meshgrid(E_y, theta_grid)
print "Length E_y"
print len(E_y)
x_points, y_points = r_points * np.cos(
theta_points), r_points * np.sin(theta_points)
x_wholeShape, y_wholeShape = r_wholeShape * np.cos(
theta_wholeShape), r_wholeShape * np.sin(theta_wholeShape)
#ax.plot_surface(x_wholeShape, y_wholeShape, wholeShape_x, rstride=1, cstride=1, cmap=cm.YlGnBu_r)
#plt.show()
#ax.plot_surface(x_points, y_points, shape_x, rstride=1, cstride=1, cmap=cm.YlGnBu_r)
#plt.show()
E_range = max(E_y) - min(E_y)
#print max(E_y)
#print min(E_y)
#N = data/E_range
#N = data/E_range
max_base = 0.0
max_trans = 0.0
max_shaft = 0.0
max_tip = 0.0
min_base = 100.0
min_trans = 100.0
min_shaft = 100.0
min_tip = 100.0
for i in range(0, len(self.youngs_modulus)):
if (self.youngs_modulus[i][0] < self.params.r_base):
if (max_base < self.youngs_modulus[i][1]):
max_base = self.youngs_modulus[i][1]
if (min_base > self.youngs_modulus[i][1]):
min_base = self.youngs_modulus[i][1]
if ((self.params.r_base < self.youngs_modulus[i][0]) &
(self.youngs_modulus[i][0] <
self.params.L - self.params.r_shaft - self.params.r_tip)):
if (max_trans < self.youngs_modulus[i][1]):
max_trans = self.youngs_modulus[i][1]
if (min_trans > self.youngs_modulus[i][1]):
min_trans = self.youngs_modulus[i][1]
if ((self.params.L - self.params.r_shaft - self.params.r_tip <
self.youngs_modulus[i][0]) &
(self.youngs_modulus[i][0] <
self.params.L - self.params.r_tip)):
if (max_shaft < self.youngs_modulus[i][1]):
max_shaft = self.youngs_modulus[i][1]
if (min_shaft > self.youngs_modulus[i][1]):
min_shaft = self.youngs_modulus[i][1]
if (self.params.L - self.params.r_tip < self.youngs_modulus[i][0]):
if (max_tip < self.youngs_modulus[i][1]):
max_tip = self.youngs_modulus[i][1]
if (min_shaft > self.youngs_modulus[i][1]):
min_tip = self.youngs_modulus[i][1]
print "Youngs Modulus maxima and minima in different regions"
print "Region\tMax\tMin"
print "Global\t" + str(max(E_y)) + "\t" + str(min(E_y))
print "Base\t" + str(max_base) + "\t" + str(min_base)
print "Neck\t" + str(max_trans) + "\t" + str(min_trans)
print "Base\t" + str(max_shaft) + "\t" + str(min_shaft)
print "Tip\t" + str(max_tip) + "\t" + str(min_tip)
N = data / 5.0
#print data
#print N
#print cm.jet(N)
#ax.plot_surface(x_wholeShape, y_wholeShape, wholeShape_x, rstride=1, cstride=1, cmap=cm.Reds)
ax.plot_surface(x_points,
y_points,
shape_x,
rstride=1,
cstride=1,
facecolors=cm.Greens_r(N),
linewidth=0,
antialiased=False,
shade=False)
#ax.set_zlim3d(0, cell_radius + shmoo_length)
#ax.set_xlabel(r'$x$')
#ax.set_ylabel(r'$y$')
#ax.set_zlabel(r'$z$')
ax.grid(False)
for a in (ax.w_xaxis, ax.w_yaxis, ax.w_zaxis):
for t in a.get_ticklines() + a.get_ticklabels():
t.set_visible(False)
a.line.set_visible(False)
a.pane.set_visible(False)
m = cm.ScalarMappable(cmap=cm.Greens_r)
m.set_array(N)
m.set_clim(0.0, 5.0)
plt.colorbar(m)
plt.show()
plt.savefig("Youngs_Modulus.png")
plt.savefig("Youngs_Modulus.eps", format="eps")
#ax.plot_surface(x_points, y_points, shape_x1, rstride=1, cstride=1, cmap=cm.YlGnBu_r)
#strain_range = max(strain_vol) - min(strain_vol)
#strain_range = max(strain_vol) - min(strain_vol)
N = data / 4.0
# print N
# print cm.jet(N)
ax.plot_surface(x_points,
y_points,
shape_x1,
rstride=1,
cstride=1,
facecolors=cm.Greens_r(N),
linewidth=1,
antialiased=False,
shade=False)
# ax.set_zlim3d(0, cell_radius + shmoo_length)
# ax.set_xlabel(r'$x$')
# ax.set_ylabel(r'$y$')
# ax.set_zlabel(r'$z$')
ax.grid(False)
for a in (ax.w_xaxis, ax.w_yaxis, ax.w_zaxis):
for t in a.get_ticklines() + a.get_ticklabels():
t.set_visible(False)
a.line.set_visible(False)
a.pane.set_visible(False)
fig = pl.figure(1, (10, 15))
ax1 = fig.add_subplot(321)
ax2 = fig.add_subplot(322)
ax3 = fig.add_subplot(323)
ax4 = fig.add_subplot(324)
ax5 = fig.add_subplot(325)
ax6 = fig.add_subplot(326)
for inh_count, inh_val in enumerate(inh_range):
ax1.plot(exc_range,
inh_exc_stn_fr[inh_count],
color=cm.Reds_r(inh_count * 30),
lw=3.,
label=str(inh_val))
ax3.plot(exc_range,
inh_exc_stn_ff[inh_count],
color=cm.Greens_r(inh_count * 30),
lw=3.,
label=str(inh_val))
ax2.plot(exc_range,
inh_exc_gpe_fr[inh_count],
color=cm.Reds_r(inh_count * 30),
lw=3.)
ax4.plot(exc_range,
inh_exc_gpe_ff[inh_count],
color=cm.Greens_r(inh_count * 30),
lw=3.)
ax5.plot(exc_range,
inh_exc_stn_spec[inh_count],
color=cm.Blues_r(inh_count * 30),
lw=3.,
label=str(inh_val))