Releases

Current release is MLatom, version 1.2.

Version 1.2

Released on 20.11.2020.

Download

Features

Manual

Tutorial

Citation:

Pavlo O. Dral, MLatom: A Program Package for Quantum Chemical Research Assisted by Machine Learning. J. Comput. Chem. 2019, 40, 2339–2347. DOI: 10.1002/jcc.26004.

Pavlo O. Dral, Bao-Xin Xue, Fuchun Ge, Yi-Fan Hou, MLatom: A Package for Atomistic Simulations with Machine Learning, version 1.2. Xiamen University, Xiamen, China, 2013–2020. http://MLatom.com.

License agreement

What’s new?

  • Change to Python 3: The python part of MLatom was upgraded to Python 3.6+.
  • New features:
    • cross section: MLatom can significantly accelerate the calculation of cross-section with the Nuclear Ensemble Approach (NEA). In brief, this feature uses fewer QC calculation to achieve higher precision and reduce computational cost. You can find more detail on this paper (please cite it when using this feature:
    • The KREG model (Kernel-ridge-regression using RE descriptor and the Gaussian kernel function) as default. Default hyperparameter optimization options were updated to get better the hyperparameters for KREG. (lgOptDepth=3, NlgSigma=6, lgLambdaL=-35)
    • minor bugs fixed

Version 1.1

Released on 30.03.2020.

Download

Features

Manual

Tutorial

Citation:

Pavlo O. Dral, MLatom: A Program Package for Quantum Chemical Research Assisted by Machine Learning. J. Comput. Chem.2019, 40, 2339–2347. DOI: 10.1002/jcc.26004.

Pavlo O. Dral, MLatom: A Package for Atomistic Simulations with Machine Learning, version 1.1. Xiamen University, Xiamen, China, 2013–2020. http://MLatom.com.

License agreement

What’s new?

  • Tutorial: Online tutorial is now available with examples how to use MLatom. It will be updated regularly based on FAQ received from you and students. It also shows examples from real studies so that the interested user can try their hand on reproducing published research.
  • New features:
    • input file: now if only one option is given to MLatom, it will consider it to be a name of the input file. Options can be given in one or several lines of the input file. Example of a typical command is: mlatom input_file.inp > output.out
    • Δ-machine learning: now you can perform Δ-machine learning [J. Chem. Theory Comput. 2015, 11, 2087] by providing MLatom with the values of the baseline method.
    • structure-based sampling from the sliced data: structure-based sampling can be now performed from data sliced into regions as described in J. Chem. Phys. 2017, 146, 244108.
    • leave-one-out cross-validation… (LOOCV) is implemented for both model evaluation and selection. LOOCV is a special type of k-fold cross-validation, when all the points are re-used for testing/validating, i.e. the number of cross-validation splits is equal to the total number of points.
      • …for model evaluation: The new keyword LOOtest requests leave-one-out cross-validation (LOOCV) for model evaluation
      • …for model selection: The new keyword LOOopt requests leave-one-out cross-validation (LOOCV) for hyperparameter optimization.
    • the exponential kernel function: The exponential kernel function can be now requested with the special option kernel=exponential.
    • debug print:
      • Printing the regression coefficients α when using the ML model.
      • Much other additional information is printed for other tasks.
  • Miscellaneous:
    • The default type of sampling keyword is set to random (in previous versions, the default type was none).
    • Self-correction option extended to any file name with reference data.

Version 1.0

Released on 19.04.2019.

Download

Features

Manual

Citation:

Pavlo O. Dral, MLatom: A Program Package for Quantum Chemical Research Assisted by Machine Learning. J. Comput. Chem.2019, 40, 2339–2347. DOI: 10.1002/jcc.26004.

Pavlo O. Dral, MLatom: A Package for Atomistic Simulations with Machine Learning, version 1.0. Max-Planck-Institut für Kohlenforschung, Mülheim an der Ruhr, Germany, 2013–2019. http://MLatom.com.

License agreement

What’s new?

  • Added wrapper python routine MLatom.py for high-level tasks
  • Much faster code for training kernel ridge regression models
  • New features:
    • permutationally invariant kernel
    • sorted variant of RE molecular descriptor (vector {req/r})
    • self-correction scheme
    • option for saving XYZ geometries with atoms sorted by nuclear repulsions
    • options for providing names of files with user-defined indices of various subsets
  • Bug fixes:
    • testing and training indices generated and saved to text files with sample sampling=random options (i.e. when no ML calculations were performed per se) were not randomly sampled. This bug did not affect ML operations.
    • In case of CVtest option and hyperparameter optimization , if the user requested to save files with indices for sub-training and validation sets, these files were generated only for one combination of cross-validation parts.

Version 0.92, revision 102

Released on 07.10.2018.

Features

Manual

Citation:

Pavlo O. Dral, MLatom: A Package for Atomistic Simulations with Machine Learning, developmental version 0.92, revision 102. Max-Planck-Institut für Kohlenforschung, Mülheim an der Ruhr, Germany, 2013–2018. http://MLatom.com.

License agreement