Features:

    - Better Separation of Concerns refactor.
    - Docs for graphical models.
    - Visualization + corner.
    - Implement LBFGS.
    - Covariance intiialization tansform search space using tchileveky decomposition of cov matrix?
    - Use importance sampling to generate initial saples with well defined priors?

Readthedocs:

    - Read docstring API, use notebook sphinx

Samples:

    - The 1D PDF array of a given parameter, via either histogram or KDE.
    - The Gaussian that best fits the PDF, via the method above, for a parameter in 1D.
    - A function that returns that highest likelihood instance of a model for a parameter between a
      range of values for that parameter.
    - parameters_for_param_index method (or use the name).

Feature grouping:

- ABC
- Graphical models.
- Model Comparison (Model comp, sensitivty maps).
- Transdimensional (phase linking, pipeline, coordinate ascent).
- Misc (Non linear search grid search).

Future:

Ideas / Other:

    - Change license LGPL
    - normal wishart distribution - prior of covariance matrix over normal distribution..
    - Prism / Linear Bayes / Emulation
    - Gaussian Processes
    - Gibbs Sampling