State of convergence 9.17.15

[pourzanj]

As of now convergence seems valid (i.e. stationarity in parameter traces) when either S is held constant or X is held constant but not when they are both moving. Perhaps there is a multimodal posterior issue going on and anchors will solve this? Having a starting point close to the generative values of Pi is also necessary for convergence.