def plot_boost(out_dir, duty_cycle=1000., max_boost=10):
"""
Generate some 3D plots for the boost.
@param out_dir: The directory to save the plots in.
@param duty_cycle: The duty cycle to use. This parameter must be a float.
@param max_boost: The max boost to use.
"""
azim = 73
elev = 28
if not os.path.exists(out_dir):
os.makedirs(out_dir)
data_path = os.path.join(out_dir, 'data.pkl')
img_path = os.path.join(out_dir, 'boost.png')
if not os.path.exists(data_path):
# Compute the range of values
duty_cycles = np.arange(0, duty_cycle + 1) / duty_cycle
min_duty_cycles = np.linspace(0, 1, duty_cycle + 1)
# Make it into a mesh
x, y = np.meshgrid(duty_cycles, min_duty_cycles)
# Evaluate the boost at each instance
z = np.array(
[[compute_boost(xii, yii, max_boost) for xii, yii in izip(xi, yi)]
for xi, yi in izip(x, y)])
with open(data_path, 'wb') as f:
cPickle.dump((x, y, z), f, cPickle.HIGHEST_PROTOCOL)
else:
with open(data_path, 'rb') as f:
x, y, z = cPickle.load(f)
# Save the plots
plot_surface(x,
y,
z,
'Active Duty Cycle',
'Minimum Active\nDuty Cycle',
'Boost',
None,
img_path,
False,
azim,
elev,
vmin=None,
vmax=None)
def make_3d_plot(p):
"""
Make a nice looking 3D plot.
@param p: The full path to the results file.
"""
# Get the data
with open(p, 'rb') as f:
sp_x, sp_y, svm_x, svm_y, param = cPickle.load(f)
sp_x, sp_y = np.median(sp_x, 1), np.median(sp_y, 1)
svm_x, svm_y = np.median(svm_x, 1), np.median(svm_y, 1)
# Fix the messed up sort order
ix = np.array(pd.DataFrame(param).sort_values([0, 1],
ascending=[True, True]).index).astype('i')
param = param[ix]
sp_x = sp_x[ix]
sp_y = sp_y[ix]
svm_x = svm_x[ix]
svm_y = svm_y[ix]
# Refactor the data
x, y, z = [], [], []
i = 0
while i < len(param):
xi, yi, zi = [], [], []
xii, yii = param[i]
p_xii = xii
while p_xii == xii:
xi.append(xii)
yi.append(yii)
zi.append(i)
i += 1
if i >= len(param):
max_overlap = yii
break
xii, yii = param[i]
x.append(xi)
y.append(yi)
z.append(zi)
x = np.array(x).T * 100
y = (np.array(y).T / max_overlap) * 100
z = np.array(z).T
# Make the plots
dir = os.path.dirname(p)
plot_surface(x, y, sp_x[z], '% Noise', '% Overlap', '\n% Error',
show=False, out_path=os.path.join(dir, 'sp_tr.png'))
plot_surface(x, y, sp_y[z], '% Noise', '% Overlap', '\n% Error',
show=False, out_path=os.path.join(dir, 'sp_te.png'))
plot_surface(x, y, svm_x[z], '% Noise', '% Overlap', '\n% Error',
show=False, out_path=os.path.join(dir, 'svm_tr.png'))
plot_surface(x, y, svm_y[z], '% Noise', '% Overlap', '\n% Error',
show=False, out_path=os.path.join(dir, 'svm_te.png'))
def make_3d_plot(p):
"""
Make a nice looking 3D plot.
@param p: The full path to the results file.
"""
# Get the data
with open(p, 'rb') as f:
sp_x, sp_y, svm_x, svm_y, param = cPickle.load(f)
sp_x, sp_y = np.median(sp_x, 1), np.median(sp_y, 1)
svm_x, svm_y = np.median(svm_x, 1), np.median(svm_y, 1)
# Fix the messed up sort order
ix = np.array(pd.DataFrame(param).sort_values([0, 1],
ascending=[True, True]).index).astype('i')
param = param[ix]
sp_x = sp_x[ix]
sp_y = sp_y[ix]
svm_x = svm_x[ix]
svm_y = svm_y[ix]
# Refactor the data
x, y, z = [], [], []
i = 0
while i < len(param):
xi, yi, zi = [], [], []
xii, yii = param[i]
p_xii = xii
while p_xii == xii:
xi.append(xii)
yi.append(yii)
zi.append(i)
i += 1
if i >= len(param):
max_overlap = yii
break
xii, yii = param[i]
x.append(xi)
y.append(yi)
z.append(zi)
x = np.array(x).T * 100
y = (np.array(y).T / max_overlap) * 100
z = np.array(z).T
# Make the plots
dir = os.path.dirname(p)
plot_surface(x, y, sp_x[z], '% Noise', '% Overlap', '\n% Error',
show=False, out_path=os.path.join(dir, 'sp_tr.png'))
plot_surface(x, y, sp_y[z], '% Noise', '% Overlap', '\n% Error',
show=False, out_path=os.path.join(dir, 'sp_te.png'))
plot_surface(x, y, svm_x[z], '% Noise', '% Overlap', '\n% Error',
show=False, out_path=os.path.join(dir, 'svm_tr.png'))
plot_surface(x, y, svm_y[z], '% Noise', '% Overlap', '\n% Error',
show=False, out_path=os.path.join(dir, 'svm_te.png'))