v0.2.0

This is the release of v0.2.0.

New features

• Add RSGCN (Renormalized Sepectral Graph Convolutional Network) (#89 thank you @anaruse!)
• Add BalancedSerialIterator for imbalanced data training (#59)
• Add ROCAUCEvaluator for binary classification task evaluation (#62)
• Add self_connection option in construct_adj_array (#100)

Changes that break compatibility

• return_smiles return numpy.ndarray, instead of list (#79)
• csv_file_parser.parse and sdf_file_parser.parse methods return dict that contains dataset and smiles, instead variable length tuple (#94)

Bug fixes

• QM9 dataset ordering (#107)

Example

• Support SchNet inference example in tox21 example (#103)
• Add RSGCN example in QM9, tox21 example (#89, #104)
• Add BalancedSerialIterator, ROCAUCEvaluator sample usage in tox21 example (#60, #62)

Others

• Dataset exploration notebook in QM9, tox21 example (#78)
• Add many tests.

Also thank you @amaotone (#50), @kazuyaujihara (#85), @msakai (#88) and @anaruse (#89) for the implementation, documentation and example improvements!