Dimod

New Features
------------
* `assert_bqm_almost_equal` function for testing
* `ScaleComposite`
* `sample_column` optional keyword argument for `SampleSet.to_pandas_df`
* `sample_dict_cast` optional keyword argument for `SampleSet.data`

Fixes
-----
* `BQM.normalize` now ignores ignored variables/interactions when determining scale

New Features
------------
* Faster `BQM.add_interaction`
* `HigherOrderComposite`
* `reverse` kwarg for `SampleSet.data`
* `BQM` alias for `BinaryQuadraticModel`
* `SampleSet.__str__` for nicer printing of samplesets

Fixes
-----
* Fix empty `SampleSet`
* Fix import warning for collections package in python 3.7
* Remove unnecessary type check in `BQM.scale`

Bug Fixes
---------
* Return dict from `BinaryQuadraticModel.to_ising` rather than `LinearView`/`QuadraticView`

New Features
-------------------
* `concatenate` function for combining sample sets
* `SampleSet.aggregate` method to aggregate repeated samples
* `SampleSet.from_samples_bqm` constructor to construct sample sets from raw samples and a binary quadratic model

Changes
------------
* Removed `OrderedBinaryQuadraticModel`
* Change the way the `BinaryQuadraticModel` stores the biases to speed up calculations and to save memory
* Deprecated `Response` in favour of `SampleSet`. `Response` is targeted to be removed in 0.9.0 release
* Updated reference samples and composites to use `SampleSet`
* Moved `dimod.io` namespace under `dimod.serialization`
* Removed `dimod.embedding` subpackage, it is now part of [dwave-system](https://github.com/dwavesystems/dwave-system) under `dwave.embedding` namespace.

New Features
-------------------
* `BinaryQuadraticModel.normalize` method, thanks wbernoudy
* `BinaryQuadraticModel.fix_variables` method
* `BinaryQuadraticModel.energies` method
* `SampleSet.to_pandas` method
* `FixedVariableComposite`, thanks conta877
* `poly_energy` and `poly_energies` function for higher order polynomials, thanks conta877
* Parameter to ignore some variables/interactions in `BinaryQuadraticModel.normalize` and `BinaryQuadraticModel.scale` methods, thanks conta877
* Fix broken documentation link, thanks schnorea

* Support empty `SampleSet` objects
* Update assert `response_energies` to work with `SampleSet`