def update_precomputed_data( self, level ):
for l in xrange( self.level+1, level+1 ):
x, w = gauss_legendre_pts_wts_1D( self.num_nodal_points( l ),
-1., 1. )
self.abscissa.append( x )
self.weights.append( w )
self.level = level
def quadrature_sub_sampling( num_dims, num_pts, max_order ):
pts = numpy.empty( ( num_dims, num_pts ), numpy.double )
I = numpy.random.randint( 0, max_order, ( num_dims, num_pts ) )
x, w = gauss_legendre_pts_wts_1D( max_order, 0., 1. )
for dim in xrange( num_dims ):
pts[dim,:] = x[I[dim,:]]
return pts
def quadrature_latin_hypercube( I ):
num_dims, num_pts = I.shape
pts = numpy.empty( ( num_dims, num_pts ), numpy.double )
x, w = gauss_legendre_pts_wts_1D( num_pts, 0., 1. )
for dim in xrange( num_dims ):
pts[dim,:] = x[I[dim,:]]
return pts
def fixed_level_quadrature_latin_hypercube( I, max_order ):
num_dims, num_pts = I.shape
I = I % max_order
I = unique_matrix_rows( I.T ).T
while ( I.shape[1] != num_pts ):
I = numpy.hstack( ( I, numpy.random.randint( 0, max_order,
( num_dims, num_pts - I.shape[1] ) ) ) )
I = unique_matrix_rows( I.T ).T
pts = numpy.empty( ( num_dims, num_pts ), numpy.double )
x, w = gauss_legendre_pts_wts_1D( max_order, 0., 1. )
for dim in xrange( num_dims ):
pts[dim,:] = x[I[dim,:]]
return pts
