Module kernel

"""
Make use of Parameter Collections to keep parameters of the
kernels. Then we would get a uniform .reset() functionality. Now reset
is provided just for parts which are failing in the unittests, but
there is many more places where they are not reset properly if
classifier gets trained on some new data of different dimensionality
"""

{'constant': KernelConstant, 'linear': KernelLinear, 'exponential': Ke
rnelExponential, 'squared exponential': KernelSquaredExponential, 'Mat
ern ni=3/2': KernelMatern_3_2, 'Matern ni=5/2': KernelMatern_5_2, 'rat
ional quadratic': KernelRationalQuadratic}