Chemml

Here is a list of new additions:

- ChemML models can now be saved and loaded [(tutorial)](https://hachmannlab.github.io/chemml/ipython_notebooks/ml_model.html#Saving-the-ChemML-model)
- The tensorflow.keras object can also be extracted from ChemML models for external use [(tutorial)](https://hachmannlab.github.io/chemml/ipython_notebooks/ml_model.html#If-the-keras-model-is-required-%E2%80%A6)
- Transfer learning can now be implemented via the new TransferLearning module using `from chemml.models import TransferLearning`
- hyperparameter optimization using genetic algorithms can now be performed using the graphical user interface (GUI) as well [(tutorial)](https://hachmannlab.github.io/chemml/ipython_notebooks/GA_tutorial.html)
- There is a small change in the installation instructions to facilitate the use of the GUI [(instructions)](https://github.com/hachmannlab/chemml#installation-and-dependencies)

Here is the list of new additions:
- local atom and bond features for graph convolutional networks
- multiprocessing for local features, coulomb matrix and bag of bonds
- multiobjective genetic algorithm

These are significant additions that pave the way for upcoming descriptors, which support multiprocessing as well >>> deserves incrementing major beta version to v0.6 >>> more is coming in v0.6.x :)

- debugged the header issue with bag_of_bonds descriptors.
- build is finally passed and tested on different platforms + added new installation instructions.

Fixed access to the static files in the datasets directory.

This is the 5th beta release so far.
The 0.4.* versions are written in Python 2 and include chemml wrapper.
The 0.5.0 version is compatible with Python 2 and 3.
The 0.5.1-2 versions are only compatible with Python 3. We decided to phase out the support for Python 2.
The 0.5.2 version provides efficient molecular descriptor implementations that only accept the built-in Molecule object as input.