visualization of criticism techniques

[dustinvtran]

for example, this includes:

• graphical checks with PPCs
• generated data alongside true data, over all dimensions
• generated data alongside true data, over a subset (or projected subset) of dimensions
• decision boundaries for classification problems

references

• stan's bayesplot package
• pymc3 plots
• seaborn examples
• bokeh gallery
• scikit-learn examples

[lbollar]

I have started some work on this. I think duplicating some of the base behavior in bayesplot would be a good idea. I have started with density plots of the posterior samples. Before I tidy up and write some docstrings, would it be best to create a separate plots folder for visualization related work?

[dustinvtran]

Awesome! Up to you; it depends on how tied they are to existing modules or if they apply across many of the existing modules. For example, the criticism folder might incorporate the visualization related work if they are interrogating the models or inference.