* No need to handle h5py datasets, compound datasets, complex datasets etc.
* Add features time permitting.
* EVERY process tool should implement two new features:
1. Check if the same process has been performed with the same paramters. When initializing the process, throw an exception. This is better than checking in the notebook stage.
2. (Gracefully) Abort and resume processing.
* Clean up Cluser results plotting
* Consider implementing doSVD as a Process. Maybe the Decomposition and Cluster classes could extend Process?
* Simplify and demystify analyis / optimize. Use parallel_compute instead of optimize and gues_methods and fit_methods
