Citrine

What's new
* This version introduces the `MeanPropertyPredictor`. This predictor computes mean ingredient properties and is meant to replace the `GeneralizedMeanPropertyPredictor`. The behavior of the `MeanPropertyPredictor` is identical to that of the `GeneralizedMeanPropertyPredictor`, but the `MeanPropertyPredictor` requires that `properties` are specified as a list of `RealDescriptor` instead of a list of descriptor keys.

Deprecated
* The `GeneralizedMeanPropertyPredictor` has been replaced by the `MeanPropertyPredictor`.

This release makes a backwards incompatible update to Alpha functionality of AutoMLPredictors, updates documentation, and clarifies some argument types.

Improvements
* In an `IngredientsToSimpleMixturePredictor`, the type of `labels` has been updated from `Mapping[str, List[str]]` to a `Mapping[str, Set[str]]`. In the `LabelFractionsPredictor`, `labels` has been updated from a `List[str]` to a `Set[str]`. (https://github.com/CitrineInformatics/citrine-python/pull/558)
* A pre-validation of `IngredientQuantityDimension` has been added to prevent the subsequent type check for the units string from failing because of a type inconsistency. (https://github.com/CitrineInformatics/citrine-python/pull/559)

Fixes
* Documentation for the `GeneralizedMeanPropertyPredictor` has been updated to reflect that `p` must be an `int`. Deprecation warnings were added in case of `float` values. (https://github.com/CitrineInformatics/citrine-python/pull/556 and https://github.com/CitrineInformatics/citrine-python/pull/560)

Deprecated
* The `AutoMLPredictor` interface has been updated from accepting `responses` as a `List[Descriptor]` to accepting `output` as a single `Descriptor`. (https://github.com/CitrineInformatics/citrine-python/pull/553)

This is a re-release of [v0.106.0](https://github.com/CitrineInformatics/citrine-python/releases/tag/v0.106.0) now that the backend platform has been updated to support this feature. This version adds a `create_default` method the `DesignSpaceCollection` that constructs a default design space from a predictor id.

What's new

- A new method `create_default` was added to the design space collection. This method will construct a default design space from a predictor id. The design space will contain dimensions and subspaces for all inputs to the predictor that cannot be computed from the predictor itself.

Improvements

- Support for units was added to `IngredientQuantityInOutput`

This release adds a `create_default` method the `DesignSpaceCollection` that constructs a default design space from a predictor id.

What's new
- A new method `create_default` was added to the design space collection. This method will construct a default design space from a predictor id. The design space will contain dimensions and subspaces for all inputs to the predictor that cannot be computed from the predictor itself.

Improvements
- Support for units was added to `IngredientQuantityInOutput`

This release introduces the `AutoMLPredictor` and an improvement to `gemd_batch_delete` functionality.

What's new
* Introducing [AutoML Predictors](https://citrineinformatics.github.io/citrine-python/workflows/predictors.html#auto-ml-predictor)! AutoML predictors allow the user fine grained control over their models and are simpler to use and understand than even the SimplePredictors. They can be linked up in a GraphPredictor to create more complex models. (https://github.com/CitrineInformatics/citrine-python/pull/545)

Improvements
* Deleting large number of items with `gemd_batch_delete` was improved. It now supports UUID-like strings and Base Entities (Objects & Templates) and paginates on groups larger than 50 elements. (https://github.com/CitrineInformatics/citrine-python/pull/546)

This release promotes Design Workflows out of alpha status, as well as the ordinary slew of quality of life improvements.

What's new
* [Design Workflows](https://citrineinformatics.github.io/citrine-python/workflows/design_workflows.html) are now beta! A Design Workflow ranks materials according to a score, providing the fundamental feedback a materials scientist needs to benefit from sequential learning. (#543)

Improvements
* The score and descriptor fields for [Design Executions](https://citrineinformatics.github.io/citrine-python/reference/citrine.resources.design_execution.html#citrine.resources.design_execution.DesignExecution) are now available through this client. (542)
* The dependencies have been updated to take advantage of the latest release of [gemd-python](https://pypi.org/project/gemd). (#544)

Fixes
* Support for Python 3.5 support was dropped in [v0.99.0](https://github.com/CitrineInformatics/citrine-python/releases/tag/v0.99.0). As this removes restriction on the version of [pandas](https://pypi.org/project/pandas/), the version used in testing has been advanced.(#544)

Deprecated
* With promotion of Design Workflows out of alpha, several associated experimental components are officially being deprecated: Cross Validation Analysis Configurations, Performance Workflow, and generic Workflows along with their Collections and Executions. (543)