swh:1:snp:d2bcff616bbf538fe8ce2a9c384200307730292a
Tip revision: af90c6e97f09f0b9a77d2fcc796f8a031ad097e8 authored by alexggmatthews on 06 June 2016, 17:06:36 UTC
Building up cone.
Building up cone.
Tip revision: af90c6e
mean_functions.py
import tensorflow as tf
import numpy as np
from .param import Param, Parameterized
class MeanFunction(Parameterized):
"""
The base mean function class.
To implement a mean funcion, write the __call__ method. This takes a
tensor X and returns a tensor m(X). In accordance with the GPflow
standard, each row of X represents one datum, and each row of Y is computed
independently for each row of X.
MeanFunction classes can have parameters, see the Linear class for an
example.
"""
def __call__(self, X):
raise NotImplementedError("Implement the __call__\
method for this mean function")
def __add__(self, other):
return Additive(self, other)
def __mul__(self, other):
return Product(self, other)
class Zero(MeanFunction):
def __call__(self, X):
return tf.zeros(tf.pack([tf.shape(X)[0], 1]), dtype='float64')
class Linear(MeanFunction):
"""
y_i = A x_i + b
"""
def __init__(self, A=np.ones((1, 1)), b=np.zeros(1)):
"""
A is a matrix which maps each element of X to Y, b is an additive
constant.
If X has N rows and D columns, and Y is intended to have Q columns,
then A must be D x Q, b must be a vector of length Q.
"""
MeanFunction.__init__(self)
self.A = Param(np.atleast_2d(A))
self.b = Param(b)
def __call__(self, X):
return tf.matmul(X, self.A) + self.b
class Constant(MeanFunction):
"""
y_i = c,,
"""
def __init__(self, c=np.zeros(1)):
MeanFunction.__init__(self)
self.c = Param(c)
def __call__(self, X):
shape = tf.pack([tf.shape(X)[0], 1])
return tf.tile(tf.reshape(self.c, (1, -1)), shape)
class Additive(MeanFunction):
def __init__(self, first_part, second_part):
MeanFunction.__init__(self)
self.add_1 = first_part
self.add_2 = second_part
def __call__(self, X):
return tf.add(self.add_1(X), self.add_2(X))
class Product(MeanFunction):
def __init__(self, first_part, second_part):
MeanFunction.__init__(self)
self.prod_1 = first_part
self.prod_2 = second_part
def __call__(self, X):
return tf.mul(self.prod_1(X), self.prod_2(X))