def as_ls(self):
return self
def as_raw_np(self):
raise ValueError, 'LibSVM calculates kernels internally; they ' +\
'cannot be converted to Numpy'
# Conversion methods
def _as_ls(kernel):
raise NotImplemented, 'LibSVM calculates kernels internally; they ' +\
'cannot be converted from Numpy'
def _as_raw_ls(kernel):
raise NotImplemented, 'LibSVM calculates kernels internally; they ' +\
'cannot be converted from Numpy'
Kernel.add_conversion('ls', _as_ls, _as_raw_ls)
class LinearLSKernel(LSKernel):
"""A simple Linear kernel: K(a,b) = a*b.T"""
__kernel_type__ = _svmc.LINEAR
__kernel_name__ = 'linear'
class RbfLSKernel(LSKernel):
"""Radial Basis Function kernel (aka Gaussian):
K(a,b) = exp(-gamma*||a-b||**2)
"""
__kernel_type__ = _svmc.RBF
__kernel_name__ = 'rbf'
gamma = Parameter(1, doc='Gamma multiplying paramater for Rbf')
'Converting data of shape %s into shogun RealFeatures'
% (data.shape,))
res = RealFeatures(data.astype(float).T)
if __debug__:
debug('KRN_SG', 'Done converting data')
return res
# Conversion methods
def _as_raw_sg(kernel):
"""Converts directly to a Shogun kernel"""
return sgk.CustomKernel(kernel.as_raw_np())
def _as_sg(kernel):
"""Converts this kernel to a Shogun-based representation"""
return PrecomputedSGKernel(matrix=kernel.as_raw_np())
Kernel.add_conversion('sg', _as_sg, _as_raw_sg)
class _BasicSGKernel(SGKernel):
"""Abstract class which can handle most shogun kernel types
Subclasses can specify new kernels using the following declarations:
- __kernel_cls__ = Shogun kernel class
- __kp_order__ = Tuple which specifies the order of kernel params.
If there is only one kernel param, this is not necessary
"""
__TODO__ = """
- Think either normalizer_* should not become proper Parameter.
"""
raise ValueError('LibSVM calculates kernels internally; they ' +\
'cannot be converted to Numpy')
# Conversion methods
def _as_ls(kernel):
raise NotImplementedError('LibSVM calculates kernels internally; '
'they cannot be converted from Numpy')
def _as_raw_ls(kernel):
raise NotImplementedError('LibSVM calculates kernels internally; they '
'cannot be converted from Numpy')
Kernel.add_conversion('ls', _as_ls, _as_raw_ls)
class LinearLSKernel(LSKernel):
"""A simple Linear kernel: K(a,b) = a*b.T"""
__kernel_type__ = _svmc.LINEAR
__kernel_name__ = 'linear'
class RbfLSKernel(LSKernel):
"""Radial Basis Function kernel (aka Gaussian):
K(a,b) = exp(-gamma*||a-b||**2)
"""
__kernel_type__ = _svmc.RBF
__kernel_name__ = 'rbf'
gamma = Parameter(1, doc='Gamma multiplying paramater for Rbf')
return res
# Conversion methods
def _as_raw_sg(kernel):
"""Converts directly to a Shogun kernel"""
return sgk.CustomKernel(kernel.as_raw_np())
def _as_sg(kernel):
"""Converts this kernel to a Shogun-based representation"""
return PrecomputedSGKernel(matrix=kernel.as_raw_np())
Kernel.add_conversion('sg', _as_sg, _as_raw_sg)
class _BasicSGKernel(SGKernel):
"""Abstract class which can handle most shogun kernel types
Subclasses can specify new kernels using the following declarations:
- __kernel_cls__ = Shogun kernel class
- __kp_order__ = Tuple which specifies the order of kernel params.
If there is only one kernel param, this is not necessary
"""
__TODO__ = """
- Think either normalizer_* should not become proper Parameter.
"""