def plot_roc(y_binary, y_binary_scores, file_name, n_classes):
pu.figure_setup()
fig_size = pu.get_fig_size(10, 8)
fig = plt.figure(figsize=fig_size)
ax = fig.add_subplot(111)
fpr = dict()
tpr = dict()
roc_auc = dict()
for i in range(n_classes):
fpr[i], tpr[i], _ = roc_curve(y_binary[:, i], y_binary_scores[:, i])
roc_auc[i] = auc(fpr[i], tpr[i])
# Compute micro-average ROC curve and ROC area
fpr["micro"], tpr["micro"], _ = roc_curve(y_binary.ravel(),
y_binary_scores.ravel())
roc_auc["micro"] = auc(fpr["micro"], tpr["micro"])
lw = 2
plt.plot(fpr["micro"],
tpr["micro"],
color='darkorange',
lw=lw,
label='ROC curve (area = %0.2f)' % roc_auc["micro"])
plt.plot([0, 1], [0, 1], color='navy', lw=lw, linestyle='--')
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('Receiver Operating Characteristic Curve')
plt.legend(loc="lower right")
pu.save_fig(fig, 'figures/roc_' + file_name + '.pdf', 'pdf')
def main(args):
# Collect data from results
df = pd.read_pickle(path='results/fixed_parameters_std-2/8.pkl')
# Set time range, 1440 = 1 day
N = 1440
x = np.arange(N)
# Prep figure
pu.figure_setup()
fig_size = pu.get_fig_size(10, 5)
fig = plt.figure(figsize=(fig_size))
############################################################################
# Collect glucose data
gut_compartment = df['gt'].values[:N]
plasma_compartment = df['mt'].values[:N]
carb_est = df['carb_ests_'].values[:N]
# carb_est[carb_est==0] = np.nan
# Plot Glucose compartments...
ax = fig.add_subplot(111)
ax.plot(x, gut_compartment, c='b', lw=pu.plot_lw(), label='Gut glucose')
ax.plot(x,
plasma_compartment,
c='r',
lw=pu.plot_lw(),
label='Plasma glucose')
ax.set_xlabel('$t \, (minutes)$')
ax.set_ylabel('$mg/L$')
ax.set_axisbelow(True)
# Add second axis
ax2 = ax.twinx()
color = 'tab:orange'
ax2.set_ylabel('$\hat{m}$', color=color)
ax2.plot(
x,
carb_est, #lw=pu.plot_lw(),
label='Carbs est.',
color=color,
alpha=.5) #, marker='o', markerfacecolor='None',
#markersize=5, linewidth=0, alpha=.2)
ax2.tick_params(axis='y', labelcolor=color)
# Add legends
ax.legend(loc='upper left', prop={'size': 5})
ax2.legend(loc='upper right', prop={'size': 5})
# Finilize plot
plt.grid()
plt.tight_layout()
if args.save:
pu.save_fig(fig, args.save)
else:
plt.show()
def main(args):
# Fetch data
df = pd.read_csv(F_NAME_DATA)
df_model = pd.read_pickle(path=F_NAME_MODEL)
# Pick a user
USR_IDS = df['user_id'].unique()
# u_id = random.choice(USR_IDS)
u_id = 8
df = df[df['user_id'] == u_id]
# Get idxs of all logged meals
df.reset_index(inplace=True)
meal_idxs = df.index[df['meal'] == 1].tolist()
df = df.iloc[meal_idxs]
df = df[df['image_url'] != '0']
glucose_len = 38
idxs = df.index.values
urls = df['image_url'].values
glucose_chunks = [
df_model['carb_ests_'].values[idx:idx + glucose_len] for idx in idxs
]
# Visualize a meal image
n = len(urls)
columns = 8
rows = max((n // 8) + 1, 2)
link = urls[0]
fig, axarr = plt.subplots(rows, columns, sharex=True, sharey=True)
for i in range(n):
img = io.imread(urls[i])
axarr[i // columns, i % columns].imshow(img)
pu.save_fig(fig, args.im_save)
plt.clf()
fig, axarr = plt.subplots(rows, columns, sharex=True, sharey=True)
for i in range(n):
axarr[i // columns, i % columns].plot(np.arange(glucose_len),
glucose_chunks[i],
c='orange')
pu.save_fig(fig, args.plt_save)
def plot_auc_binary(y_binary, y_binary_scores, file_name, n_classes):
pu.figure_setup()
fig_size = pu.get_fig_size(10, 8)
fig = plt.figure(figsize=fig_size)
ax = fig.add_subplot(111)
precision, recall, _ = precision_recall_curve(y_binary, y_binary_scores)
plt.step(recall, precision, color='b', alpha=0.2, where='post')
plt.fill_between(recall, precision, step='post', alpha=0.2, color='b')
plt.xlabel('Recall')
plt.ylabel('Precision')
plt.ylim([0.0, 1.05])
plt.xlim([0.0, 1.0])
average_precision = average_precision_score(y_binary, y_binary_scores)
plt.title('2-class Precision-Recall curve: AUC={0:0.2f}'.format(
average_precision))
pu.save_fig(fig, 'figures/auc_' + file_name + '.pdf', 'pdf')
def main(args):
data = np.random.randn(args.n_points, 2)
pu.figure_setup()
fig_size = pu.get_fig_size(10, 10)
fig = plt.figure(figsize=fig_size)
ax = fig.add_subplot(111)
ax.scatter(data[:, 0], data[:, 1], alpha=0.1, rasterized=True)
ax.set_xlabel('$x$')
ax.set_ylabel('$y$')
plt.tight_layout()
if args.save:
pu.save_fig(fig, args.save)
else:
plt.show()
def main(args):
data = np.random.randn(args.n_points, 2)
pu.figure_setup()
fig_size = pu.get_fig_size(10, 10)
fig = plt.figure(figsize=fig_size)
ax = fig.add_subplot(111)
ax.scatter(data[:, 0], data[:, 1], alpha=0.1, rasterized=True)
ax.set_xlabel('$x$')
ax.set_ylabel('$y$')
plt.tight_layout()
if args.save:
pu.save_fig(fig, args.save)
else:
plt.show()
def plot_auc(y_binary, y_binary_scores, file_name, n_classes):
pu.figure_setup()
fig_size = pu.get_fig_size(10, 8)
fig = plt.figure(figsize=fig_size)
ax = fig.add_subplot(111)
precision = dict()
recall = dict()
average_precision = dict()
for i in range(n_classes):
precision[i], recall[i], _ = precision_recall_curve(
y_binary[:, i], y_binary_scores[:, i])
average_precision[i] = average_precision_score(y_binary[:, i],
y_binary_scores[:, i])
precision["micro"], recall["micro"], _ = precision_recall_curve(
y_binary.ravel(), y_binary_scores.ravel())
average_precision["micro"] = average_precision_score(y_binary,
y_binary_scores,
average="micro")
print('Average precision score, micro-averaged over all classes: {0:0.2f}'.
format(average_precision["micro"]))
plt.step(recall['micro'],
precision['micro'],
color='b',
alpha=0.2,
where='post')
plt.fill_between(recall["micro"],
precision["micro"],
step='post',
alpha=0.2,
color='b')
plt.xlabel('Recall')
plt.ylabel('Precision')
plt.ylim([0.0, 1.05])
plt.xlim([0.0, 1.0])
plt.title(
'Average precision score, micro-averaged over all classes: AUC={0:0.2f}'
.format(average_precision["micro"]))
pu.save_fig(fig, 'figures/auc_' + file_name + '.pdf', 'pdf')
def main(args):
x = np.linspace(-6, 6, 200)
y = 1/(1 + np.exp(-x))
pu.figure_setup()
fig_size = pu.get_fig_size(10, 5)
fig = plt.figure(figsize=fig_size)
ax = fig.add_subplot(111)
ax.plot(x, y, c='b', lw=pu.plot_lw())
ax.set_xlabel('$x$')
ax.set_ylabel('$\\sigma(x)$')
ax.set_axisbelow(True)
plt.grid()
plt.tight_layout()
if args.save:
pu.save_fig(fig, args.save)
else:
plt.show()
def main(args):
x = np.linspace(-6, 6, 200)
y = 1 / (1 + np.exp(-x))
pu.figure_setup()
fig_size = pu.get_fig_size(10, 5)
fig = plt.figure(figsize=(fig_size))
ax = fig.add_subplot(111)
ax.plot(x, y, c='b', lw=pu.plot_lw())
ax.set_xlabel('$x$')
ax.set_ylabel('$\\sigma(x)$')
ax.set_axisbelow(True)
plt.grid()
plt.tight_layout()
if args.save:
pu.save_fig(fig, args.save)
else:
plt.show()
#plt.style.use('classic')
plt.figure(figsize=(4, 4))
for r in range(N_REPEATS):
plt.plot(all_results[r, :, 1], all_results[r, :, 0], '.')
plt.plot([0.0, 1.0], [0, 1.0], '--', color='k', linewidth=1.2)
#plt.arrow(0.3,0.78,0.1,0.1,-)
plt.annotate("Epoch",
xy=(0.6, 0.9),
xytext=(0.4, 0.82),
arrowprops=dict(arrowstyle="->", color="black"))
plt.xlabel("Dense Acc %")
plt.ylabel("Focused Acc %")
plt.xticks([0, .2, .4, .6, .8])
plt.yticks([0, .2, .4, .6, .8])
plt.grid(True)
save_fig("focus_acc_vs_dense")
plt.show()
# In[13]:
focus_outputs_np = np.array(focus_output).reshape((-1, HIDDEN_COUNT))
for i in range(1):
x = focus_outputs_np[:, i]
n, bins, patches = plt.hist(x, 50, normed=0, facecolor='green', alpha=0.75)
plt.xlabel('Activation Distribution')
plt.ylabel('Count')
plt.title(r'$\mathrm{Histogram\ of\ focus\ Neuron:}\ ' + str(i) + '$')
plt.grid(True)
plt.show()
import matplotlib.pyplot as plt
import plot_utils as pu
import pandas as pd
data = pd.read_csv("run-Nov06_18-21-27_juansosa-Lenovo-Legion-Y7000P-1060-tag-Loss.csv")
data = data.fillna(value=0)
data.head()
pu.figure_setup()
fig_size = pu.get_fig_size(10, 8)
print(fig_size)
fig = plt.figure(figsize=fig_size)
#fig = plt.figure()
ax = fig.add_subplot(111)
ax.plot(data['Step'], data['Value'], c='g', linewidth=0.5)
ax.set_xlabel('Steps')
ax.set_ylabel('DQN Loss')
#ax.set_axisbelow(True)
#plt.grid()
#plt.tight_layout()
pu.save_fig(fig, '1-1loss', 'pdf')
def main(args):
"""Plot multiple meals to illustrate the variance in meal responses."""
# Load data
F_NAME = 'data/minimal_tracking_data.csv'
df = pd.read_csv(F_NAME)
# Set window length of meal response
frequency = 5
w_len = (2 * 60) // frequency
x = np.arange(w_len)
# Pick a user
random.seed(42)
usr_id = random.choice(df['user_id'].unique())
df = df[df['user_id'] == usr_id]
df.reset_index(inplace=True)
# Prep data - Select frames only just after a meal is logged
meal_chunks = []
meal_idxs = df.index[df['meal'] == 1].tolist()
for meal_idx in meal_idxs:
meal_chunks.append(df.loc[meal_idx:meal_idx + w_len -
1]['quantity'].values)
# Normalize to have every line start from 0
for it, chunk in enumerate(meal_chunks):
delta = chunk[0]
meal_chunks[it] = [x - delta for x in chunk]
# Increase resolution by interpolation
x_axis = np.arange(0, w_len * frequency, frequency)
new_x_axis = np.arange(w_len * frequency) + 1
meal_chunks = [
pchip_interpolate(x_axis, chunk, new_x_axis) for chunk in meal_chunks
]
print('Numer of meals: ', len(meal_chunks))
# Prepare figure
pu.figure_setup()
fig_size = pu.get_fig_size(10, 5)
fig = plt.figure(figsize=(fig_size))
# Plot
ax = fig.add_subplot(111)
for meal in meal_chunks:
ax.plot(new_x_axis, meal, c='b', lw=.2, linestyle='dashed')
ax.set_xlabel('$t \, (minutes)$')
ax.set_ylabel('$\Delta mg/dl$')
ax.set_axisbelow(True)
plt.xlim(0, 120)
plt.grid()
plt.tight_layout()
if args.save:
pu.save_fig(fig, args.save)
else:
plt.show()
data = pd.read_csv(
"run-Dec08_17-57-30_juansosa-Lenovo-Legion-Y7000P-1060-tag-Loss_G_GAN.csv")
data = data.fillna(value=0)
data.head()
pu.figure_setup()
fig_size = pu.get_fig_size(10, 8)
print(fig_size)
fig = plt.figure(figsize=fig_size)
#fig = plt.figure()
ax = fig.add_subplot(111)
dt = data['Value'].to_numpy()
smt = smooth(dt, 0.99)
ax.plot(range(len(data['Value'])),
data['Value'],
c='orange',
linewidth=0.5,
alpha=0.3)
ax.plot(range(len(data['Value'])), smt, c='orange', linewidth=0.5)
ax.set_xlabel('Steps')
ax.set_ylabel('Loss G GAN')
#ax.set_axisbelow(True)
#plt.grid()
#plt.tight_layout()
pu.save_fig(fig, 'loss_G_GAN', 'pdf')
def main(args):
# Collect data from results
df = pd.read_pickle(path='results/fixed_parameters_std-2/4.pkl')
# Set time range
S = 540
N = 695
H = 30
x = np.arange(N - S)
# Prep figure
pu.figure_setup()
fig_size = pu.get_fig_size(10, 8)
fig = plt.figure(figsize=(fig_size))
############################################################################
# Collect CGM and meal data
cgm_data = df['cgm_inputs_'].values[S:N]
cgm_model = df['Gt'].values[S:N]
meals = df['y_meals'].values[S:N]
# Set up prediction line
pred_t = 124
pred = df['cgm_preds_'][pred_t]
x_pred = np.arange(pred_t, pred_t + len(pred))
# Make all 0 target meal NaNs
meals[meals == 0] = np.nan
# Print meal idxs
print('Meal idxs: ', np.where(meals == 1.))
# Plot CGM and Meal data
ax = fig.add_subplot(411)
ax.set_xlim(0, N - S)
ax.set_ylim(0, 250)
ax.axhline(70, linewidth=1, color='#ececec')
ax.axhline(180, linewidth=1, color='#ececec')
ax.axvspan(pred_t, pred_t + H + 1, color='#f6f6f6')
ax.plot(x,
cgm_data,
marker='o',
c='g',
linewidth=0,
markersize=1,
label='CGM')
ax.plot(x_pred,
pred,
c='black',
linewidth=1,
linestyle='dashed',
label='Modeled CGM')
ax.plot(x,
meals * 20,
marker='o',
c='purple',
linewidth=0,
markersize=4,
markerfacecolor='None',
label='Meal')
plt.vlines(x=pred_t + H,
ymin=pred[-1],
ymax=cgm_data[pred_t + H - 1],
color='r',
linewidth=2)
ax.set_axisbelow(True)
ax.legend(loc='upper left', prop={'size': 6})
ax.set_ylabel('$mg/dl$')
############################################################################
# Collect insulin data
tissue_compartment = df['s1'].values[S:N]
plasma_compartment = df['s2'].values[S:N]
bolus_data = df['bolus_inputs_'].values[S:N]
# Keep only non-zero bolus data - Set zeros to nan
bolus_data[bolus_data == 0] = np.nan
# Print bolus idxs
print('Bolus idxs: ', np.where(bolus_data > 0.))
# Plot Insulin Compartment Data
ax = fig.add_subplot(412, sharex=ax)
ax.set_xlim(0, N - S)
ax.axvspan(pred_t, pred_t + H + 1, color='#f6f6f6')
ax.plot(x,
tissue_compartment,
c='b',
lw=pu.plot_lw(),
label='Subcutaneous tissue')
ax.plot(x,
plasma_compartment,
c='r',
lw=pu.plot_lw(),
label='Blood plasma')
ax.set_ylabel('$Insulin, \, mg/L$')
# plt.grid()
ax.set_axisbelow(True)
# Add Bolus Data on second y-axis
ax2 = ax.twinx()
color = 'tab:green'
ax2.set_ylabel('Units', color=color)
ax2.plot(
x,
bolus_data, #lw=pu.plot_lw(),
label='Bolus',
color=color,
marker='D',
markerfacecolor='None',
markersize=4,
linewidth=0)
ax2.tick_params(axis='y', labelcolor=color)
ax2.set_axisbelow(True)
# Show legend
ax.legend(loc='upper left', prop={'size': 6})
ax2.legend(loc='upper right', prop={'size': 6})
############################################################################
# Collect glucose data
gut_compartment = df['gt'].values[S:N - H]
plasma_compartment = df['mt'].values[S:N - H]
xx = np.arange(N - S - H)
# Collect "prediction" data
gut_compartment_ff = df['gt'].values[N - H:N]
plasma_compartment_ff = df['mt'].values[N - H:N]
xx_ff = np.arange(N - S - H, N - S)
# Plot Glucose compartments...
ax = fig.add_subplot(413, sharex=ax)
ax.set_xlim(0, N - S)
ax.axvspan(pred_t, pred_t + H + 1, color='#f6f6f6')
ax.plot(xx, gut_compartment, c='b', lw=pu.plot_lw(), label='Gut glucose')
ax.plot(xx,
plasma_compartment,
c='r',
lw=pu.plot_lw(),
label='Plasma glucose')
ax.plot(xx_ff,
gut_compartment_ff,
c='b',
lw=pu.plot_lw(),
linestyle='dashed')
ax.plot(xx_ff,
plasma_compartment_ff,
c='r',
lw=pu.plot_lw(),
linestyle='dashed')
ax.set_ylabel('$mg/L$')
ax.set_axisbelow(True)
ax.legend(loc='upper left', prop={'size': 6})
############################################################################
# Glucose and Meal Detection
# Get glucose data
carb_est = df['carb_ests_'].values[S:N - H]
carb_mean = df['carb_mean_'].values[S:N - H]
carb_stds = df['carb_stds_'].values[S:N - H]
xx_carbs = np.arange(N - S - H)
# Get "future" glucose data
carb_est_ff = df['carb_ests_'].values[N - H:N]
xx_carbs_ff = np.arange(N - S - H, N - S)
# Get meal predictions
meal_preds = df['meal_preds_'].values[S:N - H]
meal_preds[meal_preds == 0] = np.nan
# Plot
ax = fig.add_subplot(414, sharex=ax)
ax.set_xlim(0, N - S)
ax.axvspan(pred_t, pred_t + H + 1, color='#f6f6f6')
ax.plot(xx_carbs,
carb_est,
c='orange',
lw=pu.plot_lw(),
label='Carbs est.')
ax.plot(xx_carbs_ff,
carb_est_ff,
c='orange',
lw=pu.plot_lw(),
linestyle='dashed')
# ax.plot(x, carb_mean, c='b', lw=0.5, label)
# ax.plot(xx_carbs, meal_preds, marker='^', c='magenta', linewidth=0,
# markersize=3, label='Predicted meal')
ax.legend(loc='upper left', prop={'size': 6})
ax.set_xlabel('$t \, (minutes)$')
ax.set_ylabel('$Carbs, \, mg$')
ax.set_axisbelow(True)
# plt.grid()
plt.tight_layout()
if args.save:
pu.save_fig(fig, args.save)
else:
plt.show()
TP1, FP1, TN1, FN1 = TPR_FPR_01(evaluation1, lables1, lable_dict1,thr)
TPR1 = TP1 * 1.0 / (TP1 + FN1)
FPR1 = FP1 * 1.0 / (FP1 + TN1)
p.append((TPR1 + TPR) / 2.0)
r.append((FPR1 + FPR) / 2.0)
roc += (0.5 * (FPR_0 - FPR) * TPR)
roc += (0.5 * (FPR1_0 - FPR1) * TPR1)
FPR_0 = FPR
FPR1_0 = FPR1
print(file_name, roc)
#print(p)
#print(r)
plt.plot(r, p, label=lable_names[cnt] + ', AUC=' + str(round(roc,2)))
cnt -= 1
#plt.title('')
plt.plot([0, 1], [0, 1], color='black', lw=1, linestyle='--')
plt.ylim([0., 1.05])
ax.legend(loc='lower right')
plt.ylabel('True Positive Rate')
plt.xlabel('False Positive Rate')
plt.grid()
plt.tight_layout()
pu.save_fig(fig, 'adenosquam_FPR_TPR.pdf', 'pdf')
plt.show()
def main(args):
# Collect data from results
df = pd.read_pickle(path='./../results/2019-06-15_17:48/4.pkl')
# Set time range, 1440 = 1 day
N = 3000
x = np.arange(N)
# Prep figure
pu.figure_setup()
fig_size = pu.get_fig_size(10, 50)
fig = plt.figure(figsize=(fig_size))
############################################################################
# Collect CGM and meal data
cgm_data = df['cgm_inputs_'].values[:N]
cgm_model = df['Gt'].values[:N]
meals = df['y_meals'].values[:N]
# Make all 0 target meal NaNs
meals[meals == 0] = np.nan
# Plot CGM and Meal data
ax = fig.add_subplot(411)
ax.axhspan(70, 180, alpha=0.1, color='black')
ax.set_ylim(0, 350)
ax.plot(x,
cgm_data,
marker='o',
c='g',
linewidth=0,
markersize=1,
label='CGM')
ax.plot(x,
cgm_model,
c='black',
linewidth=1,
linestyle='dashed',
label='Modeled CGM')
ax.plot(x,
meals * 20,
marker='*',
c='purple',
linewidth=0,
markersize=4,
label='Meal')
# ax.set_xlabel('$t (minutes)$')
ax.legend(loc='upper left', prop={'size': 6})
ax.set_ylabel('$mg/dl$')
meal_preds = df['meal_preds_'].values[:N]
meal_preds[meal_preds == 0] = np.nan
for it, m in enumerate(meal_preds):
if m == 1:
ax.axvline(it, c='magenta', alpha=.1)
# Add CGM prediction
# horizon = 45
# start_time = 488
# x_pred = np.arange(start_time,start_time+horizon+1)
# cgm_pred = cgm_preds[start_time]-46
# ax.plot(x_pred, cgm_pred, c='black', linewidth=1, linestyle='dashed')
############################################################################
# Collect insulin data
tissue_compartment = df['s1'].values[:N]
plasma_compartment = df['s2'].values[:N]
bolus_data = df['bolus_inputs_'].values[:N]
# Keep only non-zero bolus data - Set zeros to nan
bolus_data[bolus_data == 0] = np.nan
# Plot Insulin Compartment Data
ax = fig.add_subplot(412)
ax.plot(x,
tissue_compartment,
c='b',
lw=pu.plot_lw(),
label='Subcutaneous tissue')
ax.plot(x,
plasma_compartment,
c='r',
lw=pu.plot_lw(),
label='Blood plasma')
# ax.set_xlabel('$t \, (minutes)$')
ax.set_ylabel('$Insulin, \, mg/L$')
ax.set_axisbelow(True)
# Add Bolus Data on second y-axis
ax2 = ax.twinx()
color = 'tab:green'
ax2.set_ylabel('Units', color=color)
ax2.plot(
x,
bolus_data, #lw=pu.plot_lw(),
label='Bolus',
color=color,
marker='D',
markerfacecolor='None',
markersize=5,
linewidth=0)
ax2.tick_params(axis='y', labelcolor=color)
# Show legend
ax.legend(loc='upper left', prop={'size': 6})
# ax2.legend(loc='upper right', prop={'size': 6})
############################################################################
# Collect glucose data
gut_compartment = df['gt'].values[:N]
plasma_compartment = df['mt'].values[:N]
# Plot Glucose compartments...
ax = fig.add_subplot(413)
ax.plot(x, gut_compartment, c='b', lw=pu.plot_lw(), label='Gut glucose')
ax.plot(x,
plasma_compartment,
c='r',
lw=pu.plot_lw(),
label='Plasma glucose')
# ax.set_xlabel('$t \, (minutes)$')
ax.set_ylabel('$mg/L$')
ax.set_axisbelow(True)
ax.legend(loc='upper left', prop={'size': 6})
############################################################################
# Glucose and Meal Detection
# Get glucose data
carb_est = df['carb_ests_'].values[:N]
carb_mean = df['carb_mean_'].values[:N]
carb_stds = df['carb_stds_'].values[:N]
# Get meal predictions
meal_preds = df['meal_preds_'].values[:N]
meal_preds[meal_preds == 0] = np.nan
# Plot
ax = fig.add_subplot(414)
ax.plot(x, carb_est, c='orange', lw=pu.plot_lw())
# ax.plot(x, carb_mean, c='b', lw=pu.plot_lw(), linestyle='dashed')
# ax.plot(x, carb_mean, c='b', lw=pu.plot_lw(), linestyle='dashed')
ax.plot(x,
meal_preds * carb_est,
marker='*',
c='magenta',
linewidth=0,
markersize=4,
label='Predicted meal')
ax.set_xlabel('$t \, (minutes)$')
ax.set_ylabel('$Carbs, \, mg$')
ax.set_axisbelow(True)
plt.grid()
plt.tight_layout()
if args.save:
pu.save_fig(fig, args.save)
else:
plt.show()
"run-Dec08_17-57-30_juansosa-Lenovo-Legion-Y7000P-1060-tag-Loss_G_cycle.csv"
)
data = data.fillna(value=0)
data.head()
pu.figure_setup()
fig_size = pu.get_fig_size(10, 8)
print(fig_size)
fig = plt.figure(figsize=fig_size)
#fig = plt.figure()
ax = fig.add_subplot(111)
dt = data['Value'].to_numpy()
smt = smooth(dt, 0.99)
ax.plot(range(len(data['Value'])),
data['Value'],
c='orange',
linewidth=0.5,
alpha=0.3)
ax.plot(range(len(data['Value'])), smt, c='orange', linewidth=0.5)
ax.set_xlabel('Steps')
ax.set_ylabel('Loss G cycle')
#ax.set_axisbelow(True)
#plt.grid()
#plt.tight_layout()
pu.save_fig(fig, 'loss_G_cycle', 'pdf')
import matplotlib.pyplot as plt
import plot_utils as pu
import pandas as pd
data = pd.read_csv("run-Dec02_23-30-08_juansosa-Lenovo-Legion-Y7000P-1060-tag-Loss_supervised.csv")
data = data.fillna(value=0)
data.head()
pu.figure_setup()
fig_size = pu.get_fig_size(10, 8)
print(fig_size)
fig = plt.figure(figsize=fig_size)
#fig = plt.figure()
ax = fig.add_subplot(111)
ax.plot(data['Step'], data['Value'], c='g', linewidth=0.5)
ax.set_xlabel('Steps')
ax.set_ylabel('Supervised Loss')
#ax.set_axisbelow(True)
#plt.grid()
#plt.tight_layout()
pu.save_fig(fig, '1-2lossSupTL', 'pdf')
import matplotlib.pyplot as plt
import plot_utils as pu
import pandas as pd
data = pd.read_csv(
"run-Nov11_12-54-53_juansosa-Lenovo-Legion-Y7000P-1060-tag-Mean_Reward.csv"
)
data = data.fillna(value=0)
data2 = pd.read_csv(
"run-Dec02_23-30-08_juansosa-Lenovo-Legion-Y7000P-1060-tag-Mean_Reward.csv"
)
data2 = data2.fillna(value=0)
pu.figure_setup()
fig_size = pu.get_fig_size(10, 8)
print(fig_size)
fig = plt.figure(figsize=fig_size)
ax = fig.add_subplot(111)
ax.plot(data2['Step'], data2['Value'], c='b', linewidth=0.5, label='with TL')
ax.plot(data['Step'], data['Value'], c='g', linewidth=0.5, label='without TL')
ax.set_xlabel('Steps')
ax.set_ylabel('Mean Reward')
plt.legend()
pu.save_fig(fig, '1-2rewTL', 'pdf')
def main(args):
"""Visualize basic CGM - This can be used as a template to add
more coupled data."""
# Load data
F_NAME = 'data/minimal_tracking_data.csv'
df = pd.read_csv(F_NAME)
# Pick a user
random.seed(42)
usr_id = random.choice(df['user_id'].unique())
df = df[df['user_id'] == usr_id]
# Set dateTimetimeIndex
df['date'] = df['date'].apply(lambda x: x[:19])
df['date'] = pd.to_datetime(df['date'])
df.set_index('date', inplace=True)
# Pick a day
date = random.choice(df.index.date)
df = df[date.strftime('%Y-%m-%d')]
# Original CGM
original_x_axis = np.arange(0, 60 * 24, 5)
original_cgm_data = df['quantity']
# Increase resolution by interpolation
new_x_axis = np.arange(0, 60 * 24, 1)
new_cgm_vals = pchip_interpolate(original_x_axis, original_cgm_data,
new_x_axis)
smoothed_cgm = savgol_filter(new_cgm_vals,
window_length=15,
polyorder=1,
mode='interp')
# Remove overlap of indices
# new_x_axis = new_x_axis[~original_x_axis]
# new_cgm_vals = new_cgm_vals[~original_x_axis]
# Prepare figure
pu.figure_setup()
fig_size = pu.get_fig_size(10, 5)
fig = plt.figure(figsize=(fig_size))
# Plot
ax = fig.add_subplot(111)
# ax.axhspan(70, 180, alpha=0.1, color='#ccced1')
ax.set_ylim(125, 200)
ax.plot(new_x_axis,
smoothed_cgm,
c='black',
linewidth=1,
linestyle='dashed',
label='Smoothed data')
ax.plot(new_x_axis,
new_cgm_vals,
marker='o',
c='r',
linewidth=0,
markersize=1,
label='Upsampled data')
ax.plot(original_x_axis,
original_cgm_data,
marker='o',
c='b',
linewidth=0,
markersize=1,
label='Measured data')
ax.set_xlabel('$t \, (minutes)$')
ax.set_ylabel('$ mg/dl$')
ax.legend(loc='upper left', prop={'size': 5})
plt.xlim(0, 60 * 1 + 2)
ax.set_axisbelow(True)
plt.grid()
plt.tight_layout()
if args.save:
pu.save_fig(fig, args.save)
else:
plt.show()
def main(args):
# Collect data from results
df = pd.read_pickle(path='results/fixed_parameters_std-2/4.pkl')
# Set time range
S = 540
N = 695
H = 30
x = np.arange(N - S)
# Prep figure
pu.figure_setup()
fig_size = pu.get_fig_size(10, 5)
fig = plt.figure(figsize=(fig_size))
############################################################################
# Collect CGM and meal data
cgm_data = df['cgm_inputs_'].values[S:N]
cgm_model = df['Gt'].values[S:N]
meals = df['y_meals'].values[S:N]
# Make all 0 target meal NaNs
meals[meals == 0] = np.nan
print('Logged Meal idxs:', np.where(meals > 0))
# Get meal predictions
meal_preds = df['meal_preds_'].values[S:N]
meal_preds[meal_preds == 0] = np.nan
# Plot CGM and Meal data
ax = fig.add_subplot(211)
ax.set_xlim(0, N - S)
ax.set_ylim(0, 250)
ax.axhline(70, linewidth=1, color='#ececec')
ax.axhline(180, linewidth=1, color='#ececec')
ax.plot(x,
cgm_data,
marker='o',
c='g',
linewidth=0,
markersize=1,
label='CGM')
ax.plot(x,
meals * 20,
marker='o',
c='purple',
linewidth=0,
markersize=4,
markerfacecolor='None',
label='Meal')
ax.plot(x,
meal_preds * 5,
marker='^',
c='magenta',
linewidth=0,
markersize=4,
label='Predicted meal')
ax.set_axisbelow(True)
ax.legend(loc='upper left', prop={'size': 6})
ax.set_ylabel('$mg/dl$')
############################################################################
# Glucose and Meal Detection
# Get glucose data
carb_est = df['carb_ests_'].values[S:N]
carb_mean = df['carb_mean_'].values[S:N]
carb_stds = df['carb_stds_'].values[S:N]
# Plot
ax = fig.add_subplot(212, sharex=ax)
ax.set_xlim(0, N - S)
for it, m in enumerate(meal_preds):
if m == 1:
print('Meal prediction idx: ', it)
ax.axvline(it, c='magenta')
ax.plot(x, carb_est, c='orange', lw=pu.plot_lw(), label="$\hat{m}_t'$")
ax.plot(x, carb_mean, c='b', lw=.8, label="$\\overline{\hat{m'}}_{0:t}$")
ax.plot(x,
carb_stds,
c='b',
lw=.8,
label="$2 \cdot \sigma$",
linestyle='-.')
ax.legend(loc='upper left', prop={'size': 6})
ax.set_xlabel('$t \, (minutes)$')
ax.set_ylabel('$Carbs, \, mg$')
ax.set_axisbelow(True)
# plt.grid()
plt.tight_layout()
if args.save:
pu.save_fig(fig, args.save)
else:
plt.show()
"run-Dec08_17-57-30_juansosa-Lenovo-Legion-Y7000P-1060-tag-Loss_G_identity.csv"
)
data = data.fillna(value=0)
data.head()
pu.figure_setup()
fig_size = pu.get_fig_size(10, 8)
print(fig_size)
fig = plt.figure(figsize=fig_size)
#fig = plt.figure()
ax = fig.add_subplot(111)
dt = data['Value'].to_numpy()
smt = smooth(dt, 0.99)
ax.plot(range(len(data['Value'])),
data['Value'],
c='orange',
linewidth=0.5,
alpha=0.3)
ax.plot(range(len(data['Value'])), smt, c='orange', linewidth=0.5)
ax.set_xlabel('Steps')
ax.set_ylabel('Loss G Identity')
#ax.set_axisbelow(True)
#plt.grid()
#plt.tight_layout()
pu.save_fig(fig, 'loss_G_Iden', 'pdf')
params, q = solve_FOM(params)
return q
if __name__ == "__main__":
# choose pde [burgers,advection]
pde = "react"
case = "pacman"
Ngrid = [2**11, 2**11]
T = 1
params = params_class(case=case, pde=pde, N=Ngrid, Nt=10, T=T)
params, q = solve_FOM(params)
np.save("fft_pacman_ref_2048", q)
#%%
import matplotlib.pyplot as plt
plt.pcolormesh(params.geom.X[0], params.geom.X[1], q[..., -1])
plt.colorbar()
# %%
from plot_utils import save_fig
for it, t in enumerate(params.time.t):
k, Ek = spectrum(q[::, ::, it])
plt.loglog(k, Ek, color=[0.0, 0.0, 1 - t / params.time.T], lw=0.5)
plt.xlabel(r"wave number $|k|$")
plt.ylabel(r"energy $E(|k|)$")
plt.grid(which="both", linestyle=':')
save_fig("../imgs/energy_pacman.eps")
def main(args):
# Collect data from results
df = pd.read_pickle(path='results/fixed_parameters_std-2/8.pkl')
# Set time range, 1440 = 1 day
N = 1440
x = np.arange(N)
# Prep figure
pu.figure_setup()
fig_size = pu.get_fig_size(10, 5)
fig = plt.figure(figsize=(fig_size))
############################################################################
# Collect insulin data
tissue_compartment = df['s1'].values[:N]
plasma_compartment = df['s2'].values[:N]
bolus_data = df['bolus_inputs_'].values[:N]
# Keep only non-zero bolus data - Set zeros to nan
bolus_data[bolus_data == 0] = np.nan
# Plot Insulin Compartment Data
ax = fig.add_subplot(111)
ax.plot(x,
tissue_compartment,
c='b',
lw=pu.plot_lw(),
label='Subcutaneous tissue')
ax.plot(x,
plasma_compartment,
c='r',
lw=pu.plot_lw(),
label='Blood plasma')
ax.set_xlabel('$t \, (minutes)$')
ax.set_ylabel('$Insulin, \, mg/L$')
ax.set_axisbelow(True)
# Add Bolus Data on second y-axis
ax2 = ax.twinx()
color = 'tab:green'
ax2.set_ylabel('Units', color=color)
ax2.plot(
x,
bolus_data, #lw=pu.plot_lw(),
label='Bolus',
color=color,
marker='D',
markerfacecolor='None',
markersize=5,
linewidth=0)
ax2.tick_params(axis='y', labelcolor=color)
# Show legend
ax.legend(loc='upper left', prop={'size': 5})
ax2.legend(loc='upper right', prop={'size': 5})
plt.grid()
plt.tight_layout()
if args.save:
pu.save_fig(fig, args.save)
else:
plt.show()
import matplotlib.pyplot as plt
import plot_utils as pu
import numpy as np
from PIL import Image
from matplotlib.colors import NoNorm
pu.figure_setup()
fig_size = pu.get_fig_size(10, 8)
print(fig_size)
fig = plt.figure(figsize=fig_size)
ax = fig.add_subplot(121)
img = Image.open('../output/X_REAL/0010.png').convert('L')
arr = np.asarray(img)
plt.imshow(img, cmap='gray', norm=NoNorm())
ax.set_title('Original')
ax = fig.add_subplot(122)
img = Image.open('../output/Y/0010.png').convert('L')
arr = np.asarray(img)
plt.imshow(img, cmap='gray', norm=NoNorm())
ax.set_title('Translated')
pu.save_fig(fig, 'img2img', 'pdf')
data = pd.read_csv(
"run-Dec08_17-57-30_juansosa-Lenovo-Legion-Y7000P-1060-tag-Loss_D.csv")
data = data.fillna(value=0)
data.head()
pu.figure_setup()
fig_size = pu.get_fig_size(10, 8)
print(fig_size)
fig = plt.figure(figsize=fig_size)
#fig = plt.figure()
ax = fig.add_subplot(111)
dt = data['Value'].to_numpy()
smt = smooth(dt, 0.99)
ax.plot(range(len(data['Value'])),
data['Value'],
c='orange',
linewidth=0.5,
alpha=0.3)
ax.plot(range(len(data['Value'])), smt, c='orange', linewidth=0.5)
ax.set_xlabel('Steps')
ax.set_ylabel('Loss D')
#ax.set_axisbelow(True)
#plt.grid()
#plt.tight_layout()
pu.save_fig(fig, 'loss_D', 'pdf')
def main(args):
# Collect data from results
df = pd.read_pickle(path='results/fixed_parameters_std-2/4.pkl')
# Set time range
S = 540
N = 710
x = np.arange(N - S)
# Prep figure
pu.figure_setup()
fig_size = pu.get_fig_size(10, 5)
fig = plt.figure(figsize=(fig_size))
############################################################################
# Collect CGM and meal data
cgm_data = df['cgm_inputs_'].values[S:N]
cgm_model = df['Gt'].values[S:N]
meals = df['y_meals'].values[S:N]
# Make all 0 target meal NaNs
meals[meals == 0] = np.nan
# Print meal idxs
print('Meal idxs: ', np.where(meals == 1.))
# Plot CGM and Meal data
ax = fig.add_subplot(211)
ax.set_xlim(0, N - S)
ax.set_ylim(0, 250)
ax.axhline(70, linewidth=1, color='#e3e1e1')
ax.axhline(180, linewidth=1, color='#e3e1e1')
# ax.axvspan(125, 170, color='#eeeeee')
ax.plot(x,
cgm_data,
marker='o',
c='g',
linewidth=0,
markersize=1,
label='CGM')
ax.plot(x,
meals * 20,
marker='o',
c='purple',
linewidth=0,
markersize=4,
markerfacecolor='None',
label='Meal')
ax.set_axisbelow(True)
ax.legend(loc='upper left', prop={'size': 6})
ax.set_ylabel('$mg/dl$')
############################################################################
# Collect insulin data
tissue_compartment = df['s1'].values[S:N]
plasma_compartment = df['s2'].values[S:N]
bolus_data = df['bolus_inputs_'].values[S:N]
# Keep only non-zero bolus data - Set zeros to nan
bolus_data[bolus_data == 0] = np.nan
# Print bolus idxs
print('Bolus idxs: ', np.where(bolus_data > 0.))
# Plot Insulin Compartment Data
ax = fig.add_subplot(212, sharex=ax)
ax.set_xlim(0, N - S)
ax.plot(x,
tissue_compartment,
c='b',
lw=pu.plot_lw(),
label='Subcutaneous tissue')
ax.plot(x,
plasma_compartment,
c='r',
lw=pu.plot_lw(),
label='Blood plasma')
ax.set_xlabel('$t \, (minutes)$')
ax.set_ylabel('$Insulin, \, mg/L$')
# plt.grid()
ax.set_axisbelow(True)
# Add Bolus Data on second y-axis
ax2 = ax.twinx()
color = 'tab:green'
ax2.set_ylabel('Units', color=color)
ax2.plot(
x,
bolus_data, #lw=pu.plot_lw(),
label='Bolus',
color=color,
marker='D',
markerfacecolor='None',
markersize=4,
linewidth=0)
ax2.tick_params(axis='y', labelcolor=color)
ax2.set_axisbelow(True)
# Show legend
ax.legend(loc='upper left', prop={'size': 6})
ax2.legend(loc='upper right', prop={'size': 6})
plt.tight_layout()
if args.save:
pu.save_fig(fig, args.save)
else:
plt.show()
parsed = line_temp.split('\t')
print(parsed)
x.append(int(parsed[0]))
y.append(float(parsed[3]))
cnt += 1
return x, y
if __name__ == '__main__':
filenames = [
'AdenoSquam', 'BladderBreastLungHe', 'BladderFourBiomarkers',
'BladderFourScores', 'BreastFiveBiomarkers', 'BreastFourScores'
]
pu.figure_setup()
fig_size = pu.get_fig_size(10, 8)
fig = plt.figure(figsize=fig_size)
ax = fig.add_subplot(111)
for filename in filenames:
x, y = read_files(filename + '.txt')
plt.plot(x, y, label=filename)
ax.set_xlabel('Number of Steps')
ax.set_ylabel('Accuracy')
ax.set_axisbelow(True)
ax.legend(loc='lower right')
plt.grid()
plt.tight_layout()
pu.save_fig(fig, '../cnn_acc_steps.pdf', 'pdf')
plt.show()
import matplotlib.pyplot as plt
import plot_utils as pu
import pandas as pd
data = pd.read_csv(
"run-Nov06_18-21-27_juansosa-Lenovo-Legion-Y7000P-1060-tag-Mean_Reward.csv"
)
data = data.fillna(value=0)
data.head()
pu.figure_setup()
fig_size = pu.get_fig_size(10, 8)
print(fig_size)
fig = plt.figure(figsize=fig_size)
#fig = plt.figure()
ax = fig.add_subplot(111)
ax.plot(data['Step'], data['Value'], c='g', linewidth=0.5)
ax.set_xlabel('Steps')
ax.set_ylabel('Mean Reward')
#ax.set_axisbelow(True)
#plt.grid()
#plt.tight_layout()
pu.save_fig(fig, '1-1rew', 'pdf')
