The current code base is built off of the proof-of-concept implementation. I spent a few days refactoring it to be more sensical, but it isn't all the way there. In particular, it is half way between storing data in a global object which is referenced by the component charts and calling each of the chart functions with the current rollouts as a parameter. Half way in between is the worst of both worlds.
I prefer to have the global object in this instance since it will be easier to optimize filtering data and generating statistics if they are treated as a shared object.
Another update I would like to make is to write a better superclass for the charts so all the charts can expose a common API. This would make it easier to implement new charts following MDPvis's distribution and temporal chart formulation.
The current code base is built off of the proof-of-concept implementation. I spent a few days refactoring it to be more sensical, but it isn't all the way there. In particular, it is half way between storing data in a global object which is referenced by the component charts and calling each of the chart functions with the current rollouts as a parameter. Half way in between is the worst of both worlds.
I prefer to have the global object in this instance since it will be easier to optimize filtering data and generating statistics if they are treated as a shared object.
Another update I would like to make is to write a better superclass for the charts so all the charts can expose a common API. This would make it easier to implement new charts following MDPvis's distribution and temporal chart formulation.