Easyvvuq

New Features

* MultiEncoder element, to combine one or more encoders into a single encoder
* DirectoryEncoder element, to build directory hierarchies
* PCE Analysis element supports calculation of multiple Sobol indices
* Multi-app (Multi-solver) capabilities implemented
* Collation table may optionally be cleared using Campaign's clear_collation() method
* Added new collater (AggregateByVariables) that groups by output variable rather than run_id

Updates
* EasyVVUQ version of the Campaign and database are now compared to catch version mismatch errors
* Uniform integer distribution now available through new versions of Chaospy (thanks to collaboration with Jonathan Feinberg)
* Fix bug in collation of empty dataframes
* setup.py reads install requirements directly from requirements.txt
* Some documentation now also available as Jupyter python notebooks

v0.4-review
Intermediate release for review purposes.

New Features

* Parameter type and physical range checking (verification) implemented using Cerberus.
* Implemented a Multisampler element, allowing arbitary number of samplers to be chained together, but behave as a single sampler.
* Added a SweepSampler element, for parameter sweeps.
* Added sparse grid functionality to the Stochastic Collocation sampler.
* Added a "Worker" class (a stripped down version of the campaign) and associated tools (such as an external encoder script) to allow non-linear workflows, such as when integrating with pilot job managers.

Updates

* Improved consistency and breadth of automated testing. Campaign restarts now properly tested too.
* Added several more tutorials to the documentation.
* Fixed bug in the CSV reader with respect to column label.

Now uses current Chaospy and Scipy

New Features

* A website with [documentation](https://easyvvuq.readthedocs.io/) and a basic tutorial
* Support for multiple backend databases (via SQLAlchemy)
* Result storage in database (as opposed to in memory pandas dataframe)
* Restartable campaigns
* Incremental collation of results

Updates

* All distributions now in chaospy compatible conformations
* Varying parameters now specified in sampler rather than campaign

Feature complete version for alpha testing

Incremental development release, aligned with the M9 release of the VECMA Toolkit.

Added to the basic architecture in place for creating UQ workflows since last release:
- Chaospy integration
- Polynomial chaos expansion example
- Initial implementation of stochastic collocation