Stochastic Acceptance Step

[yannikschaelte]

• Introduce acceptor.StochasticAcceptor to encode the stochastic acceptance step replacing the standard uniform criterion.
• Introduce distance.StochasticKernel to encode noise distributions, with several concrete implementations already (Gauss, Binomial).
• Introduce epsilon.Temperature to capture the temperature replacing the traditional epsilons. In addition, multiple concrete TemperatureSchemes were implemented what handle the calculation of the next temperature value.

The goal of this implementation to be low-invasive, yet clear in the design and easily customizable. There are still a few flaws,

• e.g. that StochasticKernels are not really a subclass of Distance functions concept-wise, could be generalized to a general Comparator class.
• The information/data exchange between ABCSMC, Acceptor, Distance, Epsilon is not optimal yet and should be abstracted.

Doing these things better would however require deeper changes in the infrastructure and may thus remain for later.

Why is this no major release? I would prefer to go little steps (since this is not API breaking) and move to 0.10 series when the new functionality is mature.

Fixes #57, #58, #59, #60, #61, #192, #201

[elbaraim]

minor comments

[EmadAlamoudi]

Seems good