def get_vector_spline(self,X,Y,sigma,period): if len(self.combine_index)==self.p: spline= prob_spline.PoissonSpline(sigma = sigma,period=period) else: spline = prob_spline.MultinomialSpline(sigma = sigma,period = period) spline.fit(X,Y.T) return(spline)
def get_host_splines(self,X,Y_mat,sigma,period): splines=[] for i in range(len(Y_mat)): Y = numpy.squeeze(Y_mat[i,:]) poisson_spline = prob_spline.PoissonSpline(sigma = sigma[i], period=period) poisson_spline.fit(X, Y) splines.append(poisson_spline) return(splines)
def get_host_splines(self,X,Y_mat,sigma,period): Y = numpy.squeeze(Y_mat.T) poisson_spline = prob_spline.PoissonSpline(sigma = sigma, period=period) poisson_spline.fit(X, Y) return(poisson_spline)
s = pyplot.scatter(X,
Y,
s=30,
color='black',
zorder=3,
label='Poisson($\mu(x)$) samples')
handles.append(s)
# Build a spline using the normal loglikelihood.
normal_spline = prob_spline.NormalSpline(sigma=0)
normal_spline.fit(X, Y)
l = pyplot.plot(x,
normal_spline(x),
label='Fitted NormalSpline($\sigma =$ {:g})'.format(
normal_spline.sigma))
handles.append(l[0])
# Build a spline using the Poisson loglikelihood.
poisson_spline = prob_spline.PoissonSpline(sigma=0)
poisson_spline.fit(X, Y)
l = pyplot.plot(x,
poisson_spline(x),
label='Fitted PoissonSpline($\sigma =$ {:g})'.format(
poisson_spline.sigma))
handles.append(l[0])
# Add decorations to plot.
pyplot.xlabel('$x$')
pyplot.legend(handles, [h.get_label() for h in handles])
pyplot.show()
import seaborn
import sklearn.model_selection
import prob_spline
import test_common
npoints = 20
numpy.random.seed(2)
# Get Poisson samples around mu(x).
X_pad = 1 / 2 / npoints
X = numpy.linspace(X_pad, 1 - X_pad, npoints)
Y = scipy.stats.poisson.rvs(test_common.mu(X))
spline = prob_spline.PoissonSpline()
# Find the best sigma value by K-fold cross-validation.
kfold = sklearn.model_selection.KFold(n_splits=5, shuffle=True)
param_grid = dict(sigma=numpy.logspace(-8, 2, 21))
gridsearch = sklearn.model_selection.GridSearchCV(spline,
param_grid,
cv=kfold,
error_score=0)
gridsearch.fit(X, Y)
spline = gridsearch.best_estimator_
pyplot.subplot(2, 1, 1)
pyplot.plot(param_grid['sigma'],
gridsearch.cv_results_['mean_test_score'],
marker='o')
i = numpy.argmax(gridsearch.cv_results_['mean_test_score'])