KDEODF Class Reference

Detailed Description

ODF represented from a Kernel Density Estimate

Args:
    X (neml2.tensors.Rot): rotations, must have a single batch dimension
    kernel (Kernel): kernel function

Inheritance diagram for KDEODF:

Public Member Functions
	__init__ (self, X, kernel)
	optimize_kernel (self, miter=50, verbose=False, lr=1.0e-2)
	leave_out (self, i)
	forward (self, Y)
	__init__ (self, X, kernel)
	forward (self, Y)
	leave_out (self, i)
	optimize_kernel (self, miter=50, verbose=False, lr=0.01)
Public Member Functions inherited from ODF
	__init__ (self, X)
	n (self)
	texture_index (self, deg=5)
	__init__ (self, X)
	texture_index (self, deg=5)
	n (self)

Public Attributes
	kernel = kernel
Public Attributes inherited from ODF
	X = X

Constructor & Destructor Documentation

__init__() [1/2]

__init__	(		self,
			X,
			kernel )

__init__() [2/2]

__init__	(		self,
			X,
			kernel )

Member Function Documentation

forward() [1/2]

forward	(		self,
			Y )

Calculate the probability density at each point in Y

Args:
    Y (neml2.tensors.Rot): rotations with arbitrary batch shape

Returns:
    torch.tensor with the probabilities

forward() [2/2]

forward	(		self,
			Y )

Calculate the probability density at each point in Y

        Args:
            Y (neml2.tensors.Rot): rotations with arbitrary batch shape

        Returns:
            torch.tensor with the probabilities

leave_out() [1/2]

leave_out	(		self,
			i )

Calculate the second term of the cross-validation loss, leaving out the ith point

Args:
    i (int): index of the point to leave out

leave_out() [2/2]

leave_out	(		self,
			i )

Calculate the second term of the cross-validation loss, leaving out the ith point

        Args:
            i (int): index of the point to leave out

optimize_kernel() [1/2]

optimize_kernel	(		self,
			miter = 50,
			verbose = False,
			lr = 0.01 )

Optimize the kernel half width by cross-validation

        Keyword Args:
            miter (int): optimization iterations
            verbose (bool): if true print convergence progress
            lr (float): learning rate
            sf (float): fraction of data split out for validation

optimize_kernel() [2/2]

optimize_kernel	(		self,
			miter = 50,
			verbose = False,
			lr = 1.0e-2 )

Optimize the kernel half width by cross-validation

Keyword Args:
    miter (int): optimization iterations
    verbose (bool): if true print convergence progress
    lr (float): learning rate
    sf (float): fraction of data split out for validation

Member Data Documentation

kernel

kernel = kernel