def __init__(self, filename, exclude):
from cosmosis.datablock import option_section, names
from cosmosis.plotting.kde import KDE
exclude += ['post', 'like']
names = open(filename).readline().lstrip('#').split()
names = [[i, name] for i, name in enumerate(names) if name.split("--")[-1] not in exclude]
indices = np.array(names).T[0].astype(int)
names =np.array(names).T[1].astype(str)
n = nsample_multinest(filename)
data = np.loadtxt(filename).T
if n is not None:
print "Using last {} rows of data".format(n)
data = data[:,-n:]
if "weight" in names:
weight = data[np.argwhere(names=="weight")[0,0]]
weight /= weight.max()
else:
weight = None
indices = [index for index,name in zip(indices,names) if name!='weight' ]
names = [name for name in names if name!='weight']
self.data = data[indices]
names = [name.split("--") for name in names]
self.names = names
self.kde = KDE(self.data, weights=weight)
def smooth_likelihood(self, x, y):
n = self.options.get("n_kde", 100)
factor = self.options.get("factor_kde", 2.0)
kde = KDE([x,y], factor=factor)
x_range = (x.min(), x.max())
y_range = (y.min(), y.max())
(x_axis, y_axis), like = kde.grid_evaluate(n, [x_range, y_range])
return n, x_axis, y_axis, like
def smooth_likelihood(obj,x, y):
n = obj.options.get("n_kde", 100)
factor = obj.options.get("factor_kde", 2.0)
kde = KDE([x,y], factor=factor)
x_range = (x.min(), x.max())
y_range = (y.min(), y.max())
(x_axis, y_axis), like = kde.grid_evaluate(n, [x_range, y_range])
return n, x_axis, y_axis, like
def smooth_likelihood(obj,x, y):
n = 100
factor = 2.0
kde = KDE([x,y], factor=factor)
x_range = (x.min(), x.max())
y_range = (y.min(), y.max())
(x_axis, y_axis), like = kde.grid_evaluate(n, [x_range, y_range])
return n, x_axis, y_axis, like
class NamedKDE(object):
def __init__(self, filename, exclude):
from cosmosis.datablock import option_section, names
from cosmosis.plotting.kde import KDE
exclude += ['post', 'like']
names = open(filename).readline().lstrip('#').split()
names = [[i, name] for i, name in enumerate(names) if name.split("--")[-1] not in exclude]
indices = np.array(names).T[0].astype(int)
names =np.array(names).T[1].astype(str)
n = nsample_multinest(filename)
data = np.loadtxt(filename).T
if n is not None:
print "Using last {} rows of data".format(n)
data = data[:,-n:]
if "weight" in names:
weight = data[np.argwhere(names=="weight")[0,0]]
weight /= weight.max()
else:
weight = None
indices = [index for index,name in zip(indices,names) if name!='weight' ]
names = [name for name in names if name!='weight']
self.data = data[indices]
names = [name.split("--") for name in names]
self.names = names
self.kde = KDE(self.data, weights=weight)
def reorder_values(self, names, values):
check = np.array(self.names).T[1]
if not (check==np.array(names)).all():
#print "Will reorder parameters for interpolation."
vals = []
for name in check:
#print name
index = np.argwhere(np.array(names)==name)[0,0]
vals.append(values[index])
return vals
else:
return values
def __call__(self, values, names):
vals = self.reorder_values(names, values)
return self.kde.normalize_and_evaluate(vals)
