| 365487c | Muammar El Khatib | 05 April 2021, 17:05:37 UTC | Update neuralnetwork.py Create numpy array with dtype=object. | 05 April 2021, 17:05:37 UTC |
| 37b64c4 | Muammar El Khatib | 19 January 2021, 18:37:45 UTC | Changes to Gaussian class and handler. | 19 January 2021, 18:37:45 UTC |
| 7923396 | Muammar El Khatib | 07 January 2021, 20:47:59 UTC | Fix m2r2 issue. See https://github.com/sphinx-doc/sphinx/issues/8395 I should get rid of this dependency if the problem is not solved. | 07 January 2021, 20:47:59 UTC |
| 0b249ee | Muammar El Khatib | 18 October 2020, 18:14:40 UTC | Merge pull request #22 from vishankkumar/requirement_update Added version requirement of msgpack>=0.6.0 in requirement.txt | 18 October 2020, 18:14:40 UTC |
| a16ccba | vishankkumar | 18 October 2020, 09:42:45 UTC | Added version requirement of msgpack>=0.6.0 in requirement.txt | 18 October 2020, 09:42:45 UTC |
| cb8b816 | Muammar El Khatib | 11 October 2020, 03:07:39 UTC | m2r fix. | 11 October 2020, 03:07:39 UTC |
| b92b167 | Muammar El Khatib | 11 October 2020, 02:57:49 UTC | neuralnetwork module compatible with new pytorch versions. | 11 October 2020, 02:57:49 UTC |
| 1756320 | Muammar El Khatib | 13 September 2020, 03:13:41 UTC | Updated documentation. | 13 September 2020, 03:13:41 UTC |
| bff7ab7 | Muammar El Khatib | 13 September 2020, 02:59:48 UTC | Updated requirements for building documentation. | 13 September 2020, 02:59:48 UTC |
| bd2085b | Muammar El Khatib | 12 August 2020, 16:33:01 UTC | Modifications needed for autoencoders module work with new loss structure. | 12 August 2020, 16:33:01 UTC |
| ebbed61 | Muammar El Khatib | 10 August 2020, 04:50:14 UTC | TopologicalLoss is now a class and supports loss weights. | 10 August 2020, 04:50:14 UTC |
| 70b570e | Muammar El Khatib | 03 August 2020, 00:52:57 UTC | Merge branch 'master' of https://github.com/muammar/ml4chem into master * 'master' of https://github.com/muammar/ml4chem: Use torch.norm instead of np.linalg.norm. Addition of reconstruction to TopologicalLoss and training seems to work. Addition of Topological loss function. | 03 August 2020, 00:52:57 UTC |
| f436f77 | Muammar El Khatib | 03 August 2020, 00:16:07 UTC | Merge pull request #20 from muammar/topoae Addition of Topological loss function. | 03 August 2020, 00:16:07 UTC |
| 4c1a40e | Muammar El Khatib | 29 July 2020, 15:29:34 UTC | Use torch.norm instead of np.linalg.norm. | 29 July 2020, 15:29:34 UTC |
| 108c265 | Muammar El Khatib | 29 July 2020, 06:02:00 UTC | Addition of reconstruction to TopologicalLoss and training seems to work. | 29 July 2020, 06:02:00 UTC |
| c28fa6f | Muammar El Khatib | 27 July 2020, 02:22:22 UTC | Addition of Topological loss function. This is the starting pont of an implementation of the loss function of the paper "Topological Autoencoders" https://arxiv.org/abs/1906.00722. | 27 July 2020, 02:22:22 UTC |
| 329b12c | Muammar El Khatib | 01 June 2020, 23:51:19 UTC | Merge branch 'master' into active_learning * master: (51 commits) Support of different cutoff radius for radial/angular symmetry functions. New improvements to AEV class and documentation. Started documentation of G2 in AEV, and removed pip from readthedocs. Some more changes related to rdt. Improvements in documentations and new AEV features class. Mocking more modules and fixed typo in documentation. These changes might improve the readthedocs build. Improvements to documentation. Improving documentation. Update README.md Added ChemrXiv bibtex Fix DeepLearningModel base and rerun some of the examples. General improvements. Bump version to dev mode. Revert "Update conf.py with MagicMock" Update conf.py with MagicMock New general improvements. General improvements. `plot_atomic_features` supports now backend keyword arguments New `backend_kwargs` keyword argument for plot_atomic_features. ... | 01 June 2020, 23:51:19 UTC |
| b127fcf | Muammar El Khatib | 01 June 2020, 23:49:14 UTC | General improvements. * Pyflakes and blacked. * Added small RunTimeError for the case where non-atomistic active learning is used. * Some annotations in the source code. | 01 June 2020, 23:49:14 UTC |
| 0d14b7a | Muammar El Khatib | 19 March 2020, 05:09:07 UTC | Support of different cutoff radius for radial/angular symmetry functions. - This change is required for ANI model training and was applied to `atomistic.features.aev` and `atomistic.features.gaussian`. - Added an example and some improvements for documentation. | 19 March 2020, 05:09:07 UTC |
| 06993ed | Muammar El Khatib | 17 March 2020, 20:35:21 UTC | New improvements to AEV class and documentation. - docs/ changes to add more documentation about AEV and keep trying to fix rdt. - atomistic.features.gaussian: small change to make the class a bit more general so that it can be used by AEV. - atomistic.features.aev: addition of angular symmetry funciton. - setup.py file has an try/expect block that cheks for requirements.txt. | 17 March 2020, 20:35:21 UTC |
| 81fdf21 | Muammar El Khatib | 15 March 2020, 06:15:07 UTC | Started documentation of G2 in AEV, and removed pip from readthedocs. | 15 March 2020, 06:15:07 UTC |
| 3c702ae | Muammar El Khatib | 15 March 2020, 06:02:06 UTC | Some more changes related to rdt. | 15 March 2020, 06:02:06 UTC |
| d2dec15 | Muammar El Khatib | 15 March 2020, 05:48:53 UTC | Improvements in documentations and new AEV features class. - docs.sources.docs: * More modules are mocked for correct rendering on the rdt website. - atomistic.features.gaussian: * Changes to make the class compatible with AEV. * Change to f-strings. - atomistic.features.aev: * New AEV addition class. * It just computes G2 symmetry functions. - atomistic.modules.neuralnetwork: * When collecting the activations we create leaf variables. | 15 March 2020, 05:48:53 UTC |
| 602f987 | Muammar El Khatib | 13 March 2020, 05:23:12 UTC | Mocking more modules and fixed typo in documentation. | 13 March 2020, 05:23:12 UTC |
| 51c4145 | Muammar El Khatib | 12 March 2020, 21:13:06 UTC | These changes might improve the readthedocs build. | 12 March 2020, 21:13:06 UTC |
| 1cda502 | Muammar El Khatib | 12 March 2020, 21:07:45 UTC | Improvements to documentation. | 12 March 2020, 21:07:45 UTC |
| bf6a55e | Muammar El Khatib | 11 March 2020, 23:57:09 UTC | Improving documentation. - Updated README. - New section about HPC with ML4Chem. All this is still work in progress. | 11 March 2020, 23:57:09 UTC |
| f48c1d5 | Muammar El Khatib | 09 March 2020, 20:58:29 UTC | Update README.md | 09 March 2020, 20:58:29 UTC |
| 244df59 | Muammar El Khatib | 09 March 2020, 20:58:02 UTC | Added ChemrXiv bibtex | 09 March 2020, 20:58:02 UTC |
| 5bc7808 | Muammar El Khatib | 07 March 2020, 21:13:26 UTC | Fix DeepLearningModel base and rerun some of the examples. | 07 March 2020, 21:13:26 UTC |
| 454552c | Muammar El Khatib | 07 March 2020, 19:16:42 UTC | General improvements. - Fix metaclass issue on ml4chem.atomistic.models.base. There was a conflict resolved with `metaclass` keyword argument. - docs.source: * Updated conf.py. * Updated AE and VAE figures. | 07 March 2020, 19:16:42 UTC |
| b11e1bd | Muammar El Khatib | 06 March 2020, 18:33:17 UTC | Bump version to dev mode. | 06 March 2020, 18:33:17 UTC |
| 711087a | Muammar El Khatib | 06 March 2020, 18:33:00 UTC | Revert "Update conf.py with MagicMock" This reverts commit 1971271121b87dae1f36cbf7c7c4fd3a2a69b67a. | 06 March 2020, 18:33:00 UTC |
| 1971271 | Muammar El Khatib | 06 March 2020, 18:25:28 UTC | Update conf.py with MagicMock This might solve the problem in readthedocs. If not, changes must be reverted. | 06 March 2020, 18:25:28 UTC |
| 35b3971 | Muammar El Khatib | 06 March 2020, 17:51:08 UTC | New general improvements. - atomistic.models.neuralnetwork: New `get_activations` function to get activations of neural network. - Fixed visualization module. - Updated install site. - Bumbed version for release. | 06 March 2020, 17:51:08 UTC |
| 85681e9 | Muammar El Khatib | 28 February 2020, 22:57:30 UTC | General improvements. - Fixed load function of models. - Improved install.srt - Improvements on visualization module. | 28 February 2020, 22:57:30 UTC |
| ae54063 | Muammar El Khatib | 21 February 2020, 05:14:35 UTC | `plot_atomic_features` supports now backend keyword arguments This is helpful for preprocessing of features before doing PCA or T-SNE using `make_pipeline`. Example: ``` backend_kwargs = {"perplexity": 500} dimension = 2 dot_size = 3 plot, df = plot_atomic_features( latent_space, method="tsne", preprocessor=StandardScaler(), dimensions=dimension, backend="plotly", dot_size=dot_size, backend_kwargs=backend_kwargs ) ``` | 21 February 2020, 05:16:10 UTC |
| 8a04838 | Muammar El Khatib | 20 February 2020, 23:55:23 UTC | New `backend_kwargs` keyword argument for plot_atomic_features. | 20 February 2020, 23:55:23 UTC |
| 000aba9 | Muammar El Khatib | 19 February 2020, 23:17:06 UTC | read_log() return a dataframe instead of tuple when `data_only=True`. | 19 February 2020, 23:17:06 UTC |
| 61f8e21 | Muammar El Khatib | 18 February 2020, 18:08:42 UTC | New data_only for visualization.read_log function. | 18 February 2020, 18:08:42 UTC |
| b35ba8a | Muammar El Khatib | 17 February 2020, 05:26:02 UTC | Fix badge and documentation build at rtd. | 17 February 2020, 05:26:02 UTC |
| 31ca2d3 | Muammar El Khatib | 17 February 2020, 05:15:55 UTC | read_log() plots test error. | 17 February 2020, 05:15:55 UTC |
| fb5885c | Muammar El Khatib | 17 February 2020, 01:58:28 UTC | Support test error during training. - `NeuralNetwork` and `RetentionTime` classes support test set error prediction during training. - Black clean. - Started porting from str.format() to f"" format for better code. | 17 February 2020, 01:58:28 UTC |
| 641e4eb | Muammar El Khatib | 15 February 2020, 00:05:36 UTC | Small fix to uncertainty computation and added docs/requirements. | 15 February 2020, 00:05:36 UTC |
| 2c30d53 | Muammar El Khatib | 14 February 2020, 23:46:23 UTC | Update environment.yml probably this fixes readthedocs. | 14 February 2020, 23:46:23 UTC |
| 15ff500 | Muammar El Khatib | 13 February 2020, 22:32:05 UTC | General improvements. - Moved `Potentials` under `atomistic`. - Created new `DeepLearningTrainer` base class for the `train` class inside each model. It adds support for model checkpoint saving thanks to the `checkpoint_save` class. - `atomistic.neuralnetwork` module now supports training with uncertainties. - Small improvements to the `atomistic.models.loss` module. | 13 February 2020, 22:32:05 UTC |
| 6781f7b | Muammar El Khatib | 13 February 2020, 00:22:06 UTC | AtomicMSELoss supports uncertainty to penalize. | 13 February 2020, 00:22:06 UTC |
| 13a7aa7 | Muammar El Khatib | 12 February 2020, 19:53:15 UTC | Moved models and features into `atomistic`. - All features and models were moved to atomistic. Your scripts have to be changed to import from `ml4chem.atomistic.features` instead of `ml4chem.features` and `ml4chem.atomistic.models` instead of `ml4chem.models`. - Refactored kernelridge and gaussian_process modules. - Started update of documentation to match publication. | 12 February 2020, 19:53:15 UTC |
| 859e832 | Muammar El Khatib | 11 February 2020, 05:51:05 UTC | Update base.py Small change for the paper publication. | 11 February 2020, 05:51:05 UTC |
| 495539f | Muammar El Khatib | 31 January 2020, 19:53:10 UTC | Enforcing derived deep learning classes and general improvements. These changes are needed for ML4Chem publication. - All deep learning models are now inheriting from DeepLearningModel base class. - Visualization module moved from `data.visualization` to `.visualization`. | 31 January 2020, 19:53:10 UTC |
| 10d913a | Muammar El Khatib | 30 January 2020, 08:20:49 UTC | Bumped to 0.0.8-dev | 30 January 2020, 08:20:49 UTC |
| d490422 | Muammar El Khatib | 30 January 2020, 05:23:28 UTC | Added information when plotly does not render in Jupyter notebook. | 30 January 2020, 05:23:28 UTC |
| d4a5d60 | Muammar El Khatib | 30 January 2020, 05:03:47 UTC | Omit version of output packages in requirements.txt | 30 January 2020, 05:03:47 UTC |
| 6c3572a | Muammar El Khatib | 30 January 2020, 03:45:48 UTC | Update __init__.py Bump version. | 30 January 2020, 03:45:48 UTC |
| 0a41f49 | Muammar El Khatib | 19 January 2020, 20:29:53 UTC | Use client.submit instead of dask.compute for SVM computations Gaussian. | 19 January 2020, 20:29:53 UTC |
| 96453bc | Muammar El Khatib | 18 January 2020, 18:43:39 UTC | General improvemts - All files are pyflakes and black clean. - Improved logging for models.kernelridge, features.gaussian and features.coulombmatrix. | 18 January 2020, 18:43:39 UTC |
| 823f188 | Muammar El Khatib | 16 January 2020, 21:50:32 UTC | Improved features and models modules. - Pyflakes cleanup. - Black cleanup. - features.coulombmatrix module improved. - models.kernelridge: improved efficiency of KernelRidge regression class. | 16 January 2020, 21:50:32 UTC |
| 94b452e | Muammar El Khatib | 16 January 2020, 21:33:02 UTC | Improved efficiency of KRR. | 16 January 2020, 21:33:02 UTC |
| 0c666a2 | Muammar El Khatib | 15 January 2020, 06:37:27 UTC | Delete unused module. | 15 January 2020, 06:37:27 UTC |
| 94ff3d5 | Muammar El Khatib | 15 January 2020, 06:35:40 UTC | Renamed CoulombMatrix module and fixed KernelRidge class. - data.parser: `ani_to_ase()` function now supports list of hdf5 loaded ANI datasets. - data.features: * __init__.py: renamed CoulombMatrix module. * coulombmatrix module is more efficient when preparing SVM features, and its parameters can be saved to file. - model.kernelridge: KernelRidge class was missing a variable. - ml4chem.potentials: changed some of the functions to allow compatibility. | 15 January 2020, 06:35:40 UTC |
| bb83fd1 | Muammar El Khatib | 14 January 2020, 18:25:58 UTC | Small improvemts to features module. - features.coulomb_matrix: Computations with Futures. - features.gaussian: remove `__class__` from params file. | 14 January 2020, 18:25:58 UTC |
| 3bf7e1a | Muammar El Khatib | 14 January 2020, 06:14:07 UTC | Completed support for CoulombMartrix without preprocessing. | 14 January 2020, 06:14:07 UTC |
| 3dce822 | Muammar El Khatib | 14 January 2020, 05:55:44 UTC | Refactor the code for improving logic. | 14 January 2020, 05:55:44 UTC |
| b31a225 | Muammar El Khatib | 13 January 2020, 21:31:33 UTC | Fix Gaussian features when the preprocessor is None. | 13 January 2020, 21:31:33 UTC |
| df449da | Muammar El Khatib | 13 January 2020, 06:27:58 UTC | Updated ml4cham.features.__init__.py with CoulombMatrix | 13 January 2020, 06:27:58 UTC |
| 7f6aaf3 | Muammar El Khatib | 13 January 2020, 04:40:17 UTC | Initial support for Coulomb Matrix using DScribe. | 13 January 2020, 04:40:17 UTC |
| 906a783 | Muammar El Khatib | 13 January 2020, 01:57:32 UTC | Partial support of ANI datasets. | 13 January 2020, 02:00:40 UTC |
| 6e73579 | Muammar El Khatib | 12 January 2020, 05:01:11 UTC | Addition of `to_pandas()` method for raw ASE data and general improvements. - Improved documentation about visualizing with plotly. - Bumped version. - ml4chem.data.utils: * Addition of `ase_to_xyz()` function. - data.handler: * Addition of `to_pandas()` method. * Improved documentation. | 12 January 2020, 05:01:11 UTC |
| a041d4c | Muammar El Khatib | 07 January 2020, 22:34:14 UTC | Preparing new release. - Most of the modules print time when they were accessed. - All features now provide a .to_pandas() method. - Black cleaned. | 07 January 2020, 22:35:29 UTC |
| 91b7801 | Muammar El Khatib | 06 January 2020, 18:50:24 UTC | Merge branch 'master' into active_learning * master: General improvements. Addition of base module with abstract classes models. Improve memory usage of Gaussian() at "training", and fixed KernelRidge. Support "MultiStepLR" and "StepLR" learning rate schedulers. Refactored Gaussian() class and Potentials() can predict traj files. Addition of an Annealer class and code beautification. General improvements to ModelMerger and others. Bump for minor release. New Gaussian class changes adapted to SVM models. Gaussian class is more memory efficient for models needing tensors. New compute_mae function in metrics.py General improvements: Small improvement to README. Small change to models page and dask dashboard. | 06 January 2020, 18:50:24 UTC |
| 45e284c | Muammar El Khatib | 03 January 2020, 19:29:58 UTC | General improvements. - ml4chem.data.visualization: Added kwargs to plot_atomic_features(). - ml4chem.features: * New base class to build AtomisticFeatures.¬ * .gaussian now supports conversion to pandas DataFrames.¬ - ml4chem.models: docstrings for base module. - Black cleaned. | 03 January 2020, 19:29:58 UTC |
| d5f2cae | Muammar El Khatib | 25 December 2019, 21:47:53 UTC | Addition of base module with abstract classes models. | 25 December 2019, 21:47:53 UTC |
| 5cb63c4 | Muammar El Khatib | 16 December 2019, 23:39:29 UTC | Improve memory usage of Gaussian() at "training", and fixed KernelRidge. | 16 December 2019, 23:39:29 UTC |
| 0833137 | Muammar El Khatib | 08 December 2019, 22:54:05 UTC | Support "MultiStepLR" and "StepLR" learning rate schedulers. | 08 December 2019, 22:54:05 UTC |
| fd5f55b | Muammar El Khatib | 07 December 2019, 01:50:16 UTC | Refactored Gaussian() class and Potentials() can predict traj files. - features.gaussian: * Refactored .compute() method to allow batch computations at inference time. This makes it possible to pass an ASE trajectory file to a Potentials() class. * Black beautification. - Small improvements to preprocessing module. - models.kernelridge: * Addition of a decode() function to convert atom symbols from binary to strings. Useful function after loading from file a Kernel model in ml4chem. - models.merger: * Hardcoded Annealer when using VAE. - potentials: * 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(). TODO: . parallel energy prediction ---> model(inputs). . Implementation for SVM models. | 07 December 2019, 01:50:16 UTC |
| 1ce29b3 | Muammar El Khatib | 04 December 2019, 00:34:07 UTC | Addition of an Annealer class and code beautification. | 04 December 2019, 00:34:07 UTC |
| ad61320 | Muammar El Khatib | 03 December 2019, 22:15:36 UTC | General improvements to ModelMerger and others. - README.md: * Slack channel link. * Updated DOI. - __init__.py: bump version to 0.0.5-dev. - ModelMerger can operate with VAE. - neuralnetwork.py small logging change. | 03 December 2019, 22:15:36 UTC |
| 7dc4c26 | Muammar El Khatib | 27 November 2019, 21:36:41 UTC | Bump for minor release. | 27 November 2019, 21:36:41 UTC |
| cebb391 | Muammar El Khatib | 27 November 2019, 17:43:29 UTC | New Gaussian class changes adapted to SVM models. - Now the latest version of Gaussian class is compatible with SVM models. - data/handler.py now has the `atoms_per_image` attribute available for both training and inference. - ml4chem/utils.py get_chunks() function now should work for both SVM and non SVM models. | 27 November 2019, 17:43:29 UTC |
| b32c249 | Muammar El Khatib | 27 November 2019, 04:37:50 UTC | Gaussian class is more memory efficient for models needing tensors. | 27 November 2019, 04:37:50 UTC |
| 0d0404a | Muammar El Khatib | 21 November 2019, 23:38:35 UTC | New compute_mae function in metrics.py | 21 November 2019, 23:38:35 UTC |
| a49c7d6 | Muammar El Khatib | 21 November 2019, 22:41:53 UTC | General improvements: - Fixed examples by removing asynchronous keyword argument. - Improved in logging. - Fixed neuralnetwork.py and rt.py module for cases where batch elements are asymmetric. | 21 November 2019, 22:41:53 UTC |
| 2196266 | Muammar El Khatib | 21 November 2019, 04:56:42 UTC | Small improvement to README. | 21 November 2019, 04:56:42 UTC |
| 2434c88 | Muammar El Khatib | 21 November 2019, 00:25:52 UTC | Small change to models page and dask dashboard. | 21 November 2019, 00:25:52 UTC |
| 4d780b1 | Muammar El Khatib | 19 November 2019, 17:52:45 UTC | Merge branch 'master' into active_learning * master: (32 commits) Bump version for minor release. General improvements to autoencoders modules. VAE model now supports one_for_all keyword argument. Addition of `one_for_all` for AutoEncoder class. Renaming of modules and some classes: Moved from multivariate as bool to variant as str. Update README.md General improvements and first steps to enable VAE in ModelMerger. Improved VAE documentation and general fixes. White listed keyword arguments are initialized to None. VAE improvements. Added Multivariate Gaussian Variational Autoencoder. Fixed link on models page. General improvements. General improvements. Added interactive html plotly latent space visualization. Now plots of features have a `return`. Minor changes. Verifaction of autoencoder features consistency. Fixed missing image. ... | 19 November 2019, 17:52:45 UTC |
| e386acd | Muammar El Khatib | 15 November 2019, 19:32:50 UTC | Bump version for minor release. | 15 November 2019, 19:32:50 UTC |
| 1b4dd95 | Muammar El Khatib | 15 November 2019, 19:26:39 UTC | General improvements to autoencoders modules. - VAE reparameterize() function returns just mean values at prediction time. - Updated examples/autoencoder directory. - Renamed docs/source/ml4chem.fingerprints.rst to docs/source/ml4chem.features.rst - Potentials() class updated to properly work with new VAE changes. | 15 November 2019, 19:26:39 UTC |
| 97efd97 | Muammar El Khatib | 12 November 2019, 03:41:31 UTC | VAE model now supports one_for_all keyword argument. | 12 November 2019, 03:41:31 UTC |
| dca42d6 | Muammar El Khatib | 11 November 2019, 22:39:23 UTC | Addition of `one_for_all` for AutoEncoder class. This allows to use an unique neural network for all atom types instead of using one per each atom type as in the Behler-Parrinello scheme. In this commit it is just implemented for the vanilla AutoEncoder. | 11 November 2019, 22:39:23 UTC |
| ddca2ca | Muammar El Khatib | 10 November 2019, 21:34:08 UTC | Renaming of modules and some classes: - Renamed module fingerprints to features. - Renamed DataSet class to Data. - Renamed calculate_features method to calculate. | 10 November 2019, 21:34:08 UTC |
| cb52a0b | Muammar El Khatib | 10 November 2019, 07:06:49 UTC | Merge branch 'master' of https://github.com/muammar/ml4chem * 'master' of https://github.com/muammar/ml4chem: Update README.md | 10 November 2019, 07:06:49 UTC |
| 22de5fb | Muammar El Khatib | 10 November 2019, 07:05:14 UTC | Moved from multivariate as bool to variant as str. With these changes it becomes possible to support the following VAE variants: - "multivariate": decoder outputs a distribution with mean and variance, we minimize the negative of the log likelihood plus the KL-Divergence. Useful for continuous variables. Feature range [-inf, inf]. - "bernoulli": decoder outputs a layer with sigmoid activation function, and we minimize cross-entropy plus KL-diverence. Features must be in a range [0, 1]. - "dcgan": decoder outputs a single layer with tanh, and loss equals to KL-Diverngence plus MSELoss. Useful for feature ranges [-1, 1]. | 10 November 2019, 07:05:14 UTC |
| 14c3d05 | Muammar El Khatib | 09 November 2019, 18:18:48 UTC | Update README.md | 09 November 2019, 18:18:48 UTC |
| 02e5a43 | Muammar El Khatib | 07 November 2019, 01:21:14 UTC | General improvements and first steps to enable VAE in ModelMerger. | 07 November 2019, 01:21:14 UTC |
| 1028721 | Muammar El Khatib | 06 November 2019, 18:42:33 UTC | Improved VAE documentation and general fixes. - Added improved VAE documentation. - Now the VAELoss function is working as expected. - I passed `black` to the whole code base. | 06 November 2019, 18:42:33 UTC |
| f7cb1c9 | Muammar El Khatib | 06 November 2019, 06:52:04 UTC | White listed keyword arguments are initialized to None. If white listed keyword arguments are not passed by users we initialized them to None. | 06 November 2019, 06:52:04 UTC |
| b5c9fc0 | Muammar El Khatib | 06 November 2019, 05:27:15 UTC | VAE improvements. | 06 November 2019, 05:27:15 UTC |
| 93f3df0 | Muammar El Khatib | 06 November 2019, 05:24:32 UTC | Added Multivariate Gaussian Variational Autoencoder. - Instantiate VAE with `multivariate=True`. | 06 November 2019, 05:24:32 UTC |
| 51d32d4 | Muammar El Khatib | 01 November 2019, 05:31:58 UTC | Fixed link on models page. | 01 November 2019, 05:31:58 UTC |
| e026b11 | Muammar El Khatib | 01 November 2019, 05:19:19 UTC | General improvements. - VAE is now implemented as shown in https://keras.io/examples/variational_autoencoder/ - Improved documentation. - Change dot_size in ml4chem/data/visualization.py. - Reconstruction loss VAE is now multiplied by input dimension. | 01 November 2019, 05:19:19 UTC |
