Source code for infpy.gp.neural_network_kernel

#
# Copyright John Reid 2006, 2012
#


from kernel import *
from real_kernel import *
import numpy, math

class NeuralNetworkKernel( RealKernel ):
    """Following 4.29 in `Gaussian Processes for Machine Learning`__ by Rasmussen and Williams. 

    __ http://www.amazon.co.uk/Gaussian-Processes-Learning-Adaptive-Computation/dp/026218253X/

    No trainable paramaters
    """

    def __init__(
            self,
            sigma
    ):
        self.sigma = numpy.matrix( sigma, numpy.float64 )
        if self.sigma.shape[0] != self.sigma.shape[1]:
            raise RuntimeError( """Sigma must be a square matrix of size equal to """
                    """dimension of the data: %d, %d""" % self.sigma.shape )
        RealKernel.__init__(
                self,
                numpy.asarray([ ]),
                [ ],
                self.sigma.shape[0]
        )

    def _calc_term( self, x1, x2 ):
        return 2.0 * x1 * self.sigma * x2.T

    def __str__( self ):
        return """NeuralNetworkKernel"""

    def __call__( self, x1, x2, identical = False ):
        (x1, x2) = self._check_args( x1, x2 )
        x1 = numpy.matrix( x1 )
        x2 = numpy.matrix( x2 )
        return (
                2.0 / math.pi
                * math.asin(
                        self._calc_term( x1, x2 )
                        / math.sqrt(
                                self._calc_term( x1, x1 )
                                * self._calc_term( x2, x2 )
                        )
                )
        )