https://github.com/muammar/ml4chem - 0.0.9

Fix DeepLearningModel base and rerun some of the examples.

- Python
Published by muammar almost 6 years ago

https://github.com/muammar/ml4chem - 0.0.8

These changes are related to ML4Chem’s publication:

• Creation of atomistic module to comply with publication structure.

• models:

  • All models comply with base class inheritance using mixin.
  • models moved to atomistic.models. Scripts have to be changed as the following example from ml4chem.atomistic.models.neuralnetwork import NeuralNetwork.
  • SVM model refactoring.
  • training class can take checkpoint dictionary to save checkpoints during training.
  • NeuralNetwork and RetentionTime models support training with test error.
  • New get_activations function to get activations of neural network.
    

• features:

  • features moved to atomistic.features. Scripts have to be changed as the following example from ml4chem.atomistic.features.gaussian import Gaussian.

• data.visualization moved to .visualization.

  • read_log() function now supports data_only keyword argument to return only data instead of matplotlib plot.
  • plot_atomic_features supports now backend keyword arguments and preprocessing of features before doing PCA or T-SNE using make_pipeline. Example:

``` backendkwargs = {"perplexity": 500} dimension = 2 dotsize = 3

plot, df = plotatomicfeatures( latentspace, method="tsne", preprocessor=StandardScaler(), dimensions=dimension, backend="plotly", dotsize=dotsize, backendkwargs=backend_kwargs ) ```

• Restructured documentation.

• AtomicMSELoss supports penalization using a list of uncertainties.

• Potentials class moved to atomistic module. It has to be imported as from ml4chem.atomistic import Potentials.

- Python
Published by muammar almost 6 years ago

https://github.com/muammar/ml4chem - 0.0.7

Omit version of output packages in requirements.txt

- Python
Published by muammar about 6 years ago

https://github.com/muammar/ml4chem - 0.0.6

• data.parser:
  • Support ANI datasets.
  • Renamed FakeCalculator to SinglePointCalculator.
• features.gaussian: Refactored Gaussian class.
• Support of Coulomb Matrix with DScribe. This serves as an example to implement all descriptors available in that library.
• models.kernelridge partially improved efficiency of KernelRidge class.

- Python
Published by muammar about 6 years ago

https://github.com/muammar/ml4chem - 0.0.5

• New to_pandas() method to convert features to DataFrame.
• Print date where a module was accessed.
• New base classes that can be used to build new features and model modules.
• New Annealer() class for training VAEs. Right now it is hardcoded in the VAE class but will be improved later.
• Addition of MultiStepLR and StepLR learning rate schedulers.
• ml4chem.data.visualization: Added kwargs to plotatomicfeatures()
• Improved memory usage of Gaussian() at "training", and fixed KernelRidge.
• A batch_size keyword argument can be passed to the Potentials.load() function so that we can do predictions of trajectory files instead of Atoms().

- Python
Published by muammar about 6 years ago

https://github.com/muammar/ml4chem - 0.0.4

Changelog:

• Added new compute_mae() function to ml4chem.metrics.
• Improved memory management on Gaussian() class.

- Python
Published by muammar about 6 years ago

https://github.com/muammar/ml4chem - 0.0.3

Changelog:

• Interactive plotting support with addition of plotly.
• Improved documentation.
• Addition of a Variational Autoencoder class (VAE), and a VAELoss loss function.
• Renamed module fingerprints to features.
• Renamed class DataSet to Data.

- Python
Published by muammar over 6 years ago

https://github.com/muammar/ml4chem - 0.0.2

Changelog:

• Improvements in the documentation.
• Better memory management in Gaussian class.
• Implement Weighted ACSFs. Issue: https://github.com/muammar/ml4chem/issues/8.
• Change in Header.
• ModelMerger class works with both dependent and independent loss functions. Issue: https://github.com/muammar/ml4chem/issues/10.
• Added install_requires to setup.py.
• CJson parser. Issue: https://github.com/muammar/ml4chem/issues/11.

- Python
Published by muammar over 6 years ago

https://github.com/muammar/ml4chem - 0.0.1

Initial release! The package still is under heavy development and is expected to mutate massively during the following releases.

- Python
Published by muammar over 6 years ago