Releases | Open Source Science

ECNet - alvaDesc bug fix

Sets alvaDesc na\r return values to 0.0

Scientific Software - Peer-reviewed - Python
Published by tjkessler almost 2 years ago

ECNet - Update dependencies

update all package dependencies

Scientific Software - Peer-reviewed - Python
Published by tjkessler almost 2 years ago

ECNet - 4.1.2 - Update build/install method, add GitHub workflows, unittest -> pytest

Build/installation now uses pyproject.toml instead of the deprecated setup.py
Added GitHub workflows for PyPI publishing and unit testing
Unit tests now use pytest instead of unittest

Scientific Software - Peer-reviewed - Python
Published by tjkessler almost 3 years ago

ECNet - Dependency update

Update to package dependencies, notably PyTorch 1.8.0 -> 2.0.0. ECNet now requires Python 3.11+.

Scientific Software - Peer-reviewed - Python
Published by tjkessler about 3 years ago

ECNet - Updates to training runtime, tuning arguments

Added option to shuffle training/validation subsets every epoch
Update to docstrings/documentation
Added a "getting started" notebook in the examples directory
New argument format for ABC-based parameter tuning

Scientific Software - Peer-reviewed - Python
Published by tjkessler almost 5 years ago

ECNet - PyTorch rework, new API, bundled property sets

ECNet now leverages the PyTorch package for ML operations
- This change presented an opportunity to overhaul ECNet from the ground up, allowing us to think about how the user will interact with this package. Ultimately, we wanted to make interactions easier.
Custom data structures were weird, and didn't belong in a ML toolkit. Instead, we offer PyTorch-based data structures, adjusted to house chemical data. Users can obtain SMILES strings and property values, or a ML-ready structure ready to be passed to ECNet for training.
All these changes require documentation, so full API documentation is available. We also have an example script, and would like to include more examples in the future.

Scientific Software - Peer-reviewed - Python
Published by tjkessler about 5 years ago

ECNet - Update to ECabc-based hyper-parameter tuning functions

Per ECabc's API changes in its 3.0.0 update, this ECNet update incorporates these changes into all relevant functions.

Scientific Software - Peer-reviewed - Python
Published by tjkessler over 5 years ago

ECNet - Better implementation of TensorFlow 2.0

ecnet.models.mlp.MultilayerPerceptron's implementation now makes sense, and leads to faster training times
some database cleanup
in case ECNet is not utilized to use a pre-trained project, input QSPR descriptor names are also saved inside the project (DataFrame object not required)

Scientific Software - Peer-reviewed - Python
Published by tjkessler over 6 years ago

ECNet - ML back-end, workflows, database additions/encoding, and more

Addition of validated, PaDEL/alvaDesc-generated YSI databases
Update to repository links, author information
ecnet.utils.data_utils now forces UTF-8 encoding for all database creation/saving
ML back-end updated to TensorFlow 2.0.0
- No API changes to ecnet.models.mlp.MultilayerPerceptron
- Existing .h5 model files will not work with the updated class
Note: initially, PyTorch was looked at as an alternative; however, after tests to evaluate performance were conducted and the viability of installing PyTorch on high-performance machines available to the ECRL were both deemed inadequate, updating to TensorFlow 2.0.0 was deemed the most appropriate action.
Only the following hyper-parameters are tuned with the built-in functions:
- Learning rate of Adam optimization function
- Learning rate decay of Adam optimization function
- Batch size during training
- Patience (if validating, # epochs to wait for better validation loss, else terminate training)
- Size of each hidden layer
Note: with the relatively small number of samples our models are trained with, it does not make sense to adjust hyper-parameters such as **beta1, **beta_2, and epsilon. The hyper-parameters listed above are theorized to play a much more important role with how the models train/perform._
Added the UML ECRL's general publication workflow as ecnet.workflows.ecrlworkflow.createmodel
If using ecnet.Server and not creating a project, a single model's filename can now be specified as an additional argument (default: model.h5)
TensorFlow's verbose argument is now propagated from ecnet.Server.train to the model during training; added as an additional argument
ecnet.models.mlp.MultilayerPerceptron.fit now returns a tuple: (learn losses, validation losses); learn losses and validation losses are both lists containing loss values (mean squared error) at every epoch; if training a single model using ecnet.Server.train, this tuple is returned; if not performing validation, the validation losses list is populated with None elements equal in size to the learn losses list
If installing using setup.py, installing TensorFlow is optional; to skip the installation of the pre-compiled PyPI distribution of TensorFlow, run setup.py with python setup.py --omit_tf install

Note: other methods of installing TensorFlow offer clear benefits (GPU support, different CPU instruction sets, etc.), therefore we want to provide an option for the user to use an existing installation of TensorFlow instead of forcing the PyPI-sourced version.

Scientific Software - Peer-reviewed - Python
Published by tjkessler over 6 years ago

ECNet - Bug fixes

If validation/test sets are empty, input parameter limiting processes will still run
Server.limit_inputs now correctly returns input parameter names, importances

Scientific Software - Peer-reviewed - Python
Published by tjkessler almost 7 years ago

ECNet - Bug fixes, enhancements

ecnet.Server.removeoutliers_ and ecnet.tasks.removeoutliers_ have been removed
- while detecting outliers may be beneficial in determining abnormalities in data, removing them entirely is likely not the right approach (in terms of fuel property prediction). Once a viable usage has been determined, outlier detection will be included.
Added the batchsize_ hyper-parameter, included in the default model configuration and hyper-parameter tuning process
- Relevant unit tests updated
Any missing model configuration variables from config files generated with previous versions of ECNet will now be set to their default values
- Additional unit tests added
Added option to convert SMILES to MDL during PaDEL-based database creation
- Additional unit test added
Added PaDEL-generated databases for all properties
ecnet.tasks.limit_inputs.limit_rforest now relies on sklearn.ensemble.RandomForestRegressor as its only dependency
- limit_rforest now returns list of parameter names/importances instead of a modified DataFrame
- Server.limit_inputs also returns a list of parameter names/importances
- Removed the ditto-lib dependency
Bug fixes:
- Server._sets now loads when a PRJ file is opened via ecnet.Server
- ecnet.utils.data_utils.DataFrame.set_inputs now immediately applies selected inputs to L/V/T sets
- ParityPlot parity lines now scale to reflect data minimum/maximum
More robust unit tests for MultilayerPerceptron, database creation, input parameter limiting
All unit tests may now be run individually

Scientific Software - Peer-reviewed - Python
Published by tjkessler almost 7 years ago

ECNet - Better MLP validation, moved multiprocessing checks

Training an MLP using a validation set now uses Keras' early stopping callback to determine learning cutoff, preserves weights at best validation loss
Moved multiprocessing.setstartmethod to multiprocessed tasks

Scientific Software - Peer-reviewed - Python
Published by tjkessler almost 7 years ago

ECNet - Removal of conversion functions, slight Server rework

1.) The following conversions have been removed from ECNet:

get_smiles
smilestodescriptors
smilestomdl
mdltodescriptors

*Note: these were adding clutter, and were not within the main scope of ECNet.

2.) PaDEL-Descriptor is no longer bundled into ECNet

*Note: with the removal of conversion functions, this is no longer needed.

3.) Database creation functions now rely on two separate packages:

PaDELPy (https://github.com/ECRL/PaDELPy) - QSPR descriptor generation using PaDEL-Descriptor
alvaDescPy (https://github.com/ECRL/alvaDescPy) - QSPR descriptor generation using alvaDesc

*Note: it made sense to create separate packages for interfacing with these software, a Python interface for generating QSPR descriptors is generally quite handy.

4.) ecnet.tools.database.createdb_'s arguments have been changed:

```python

ecnet.tools.database.createdb(['CC', 'CCC'], 'mydatabase.csv', targets=[13, 47]) ```

Construct using alvaDesc:

```python

ecnet.tools.database.createdb(['CC', 'CCC'], 'mydatabase.csv', targets=[13, 47], backend='alvadesc') ```

*Note: supplying SMILES strings and targets using lists makes more sense than requiring the user to create a separate file - this change allows the user to choose where the data comes from.

5.) ecnet.tools.project.predict's arguments have been changed:

```python

results = ecnet.tools.project.predict(['CC', 'CCC'], 'my_project.prj') print(results) [[13], [47]] ```

*Note: similar to why we switched to lists as inputs in database creation, makes more sense

6.) ecnet.Server has been rearranged a bit:

project training has been moved to a separate function at ecnet.tasks.training.trainproject_
various functions have been moved to ecnet.utils.serverutils_:
- creating a project folder structure
- saving a project as a .prj file
- opening a .prj file to use
task-specific logging messages have been moved to their respective functions in ecnet.tasks

*Note: ecnet.Server needed to be shrunk down, and functions that were obviously utilities were moved into utility files. This should also provide more direct access to the "back-end" of ECNet (subverting Server usage), allowing greater variation in experimental procedure.

7.) Added a suite of unit tests implemented with the unittest library:

in addition to Server unit tests, individual utilities of ECNet are tested
added a Python script, /tests/testall.py_, to automatically run all unit tests and report a summary of successes/failures

*Note: it's time for "proper" unit testing, and that means implementing a unit testing package. I'm looking forward to expanding ECNet's tests and introduce more automation into the testing process.

8.) Installation now forces TensorFlow 1.13.1 to be installed

*Note: I've encountered pip install tensorflow installing the 2.0.0 beta, which ECNet does not currently support - we'll make the change when we're ready (and so is Keras)

9.) Changed/added a variety of databases to the /databases/ directory

All databases constructed using alvaDesc
All SMILES strings have been validated with respect to compound name
- PubChemPy (https://github.com/mcs07/PubChemPy) is a lifesaver
- Compounds not found on PubChem were validated in-house by an ECRL research assistant

*Note: in order to ensure accurate QSPR-descriptor to experimental value correlation, accurate SMILES strings are necessary (assuming descriptors are being generated using them).

Scientific Software - Peer-reviewed - Python
Published by tjkessler about 7 years ago

ECNet - Type checking, improved unit testing

All methods/functions now enforce specific types for arguments, return values
calcr2 function now uses scikit-learn's r2score function
Changed unit testing scheme, now uses unittest library
- added a suite of unit tests

Scientific Software - Peer-reviewed - Python
Published by tjkessler about 7 years ago

ECNet - Addition to conversion tools, update to database creation function

Addition of the "smilestodescriptors" function
Database creation functions now use the "smilestodescriptors" function, bypassing the use of OpenBabel (used for SMILES -> MDL -> descriptors)
Updated relevant documentation

Scientific Software - Peer-reviewed - Python
Published by tjkessler about 7 years ago

ECNet - Updates to DataFrame, DataPoint classes and their functionality

STRING and GROUP attributes for DataPoints (rows in an ECNet-formatted database) can now be accessed as object attributes. For example:

```python

from ecnet.utils.datautils import DataFrame df = DataFrame('mydatabase.csv') firstentry = df.datapoints[0] print(firstentry.SMILES) # SMILES is a STRING column in the supplied database C print(getattr(firstentry, 'Compound Name') # STRINGs with spaces are obtained like this Methane ```

Additional STRING columns can be supplied when creating an ECNet-formatted database
Fixed issue where YAML package was throwing a loader warning
Suppressed TensorFlow warnings about deprecation
Updates to documentation
Other minor changes

Scientific Software - Peer-reviewed - Python
Published by tjkessler about 7 years ago

ECNet - Bug fixes, database creation improvements

Added "setspawnmethod", fixes multiprocessing on Unix systems
Databases can now be constructed with fingerprints instead of descriptors
"get_smiles" function now returns an empty string if the molecule is not found on PubChem
Slight updates to logging
Hyperparameter tuning bug fix

Scientific Software - Peer-reviewed - Python
Published by tjkessler about 7 years ago

ECNet - 3.0.0 Release

Server object refactor
- Includes API changes
Update to ML back end (raw TensorFlow -> Keras)
Logging moved to separate module
Input descriptor limiting now uses random forest regression, via ditto-lib 1.0.0
Implemented ReadTheDocs page
Added classes for parity plot generation
Updated hyperparameter tuning for ECabc 2.2.2 release
Implemented methods for removing outliers, via ditto-lib 1.0.0

Scientific Software - Peer-reviewed - Python
Published by tjkessler about 7 years ago

ECNet - Bug fixes, GA improvements, data sorting options, optimizations

Updated parameter limiting with GA, per 0.6.0 PyGenetics update
Fixed bug with MultilayerPerceptron returning "NaN" values
Changed default parameter bounds for ABC tuning
Added "sortstring" argument for Server.importdata
Added "Getting Started" tutorial for new users

Scientific Software - Peer-reviewed - Python
Published by tjkessler over 7 years ago

ECNet - Tool integrations, database additions

Integrated various tools:
- Database creation tool (wrappers for Open Babel, PaDEL-Descriptor)
- Using project tool (supply text w/ molecules, ECNet .prj file)
- Get SMILES from molecule name (PubChemPy)
- Convert SMILES to MDL/SDF
- Convert MDL/SDF to QSPR descriptors
Added unit tests for database creation tool, using project tool
Removed command line tools (integrated, above)
Added various databases:
- Cloud point
- Pour point
- Yield sooting index

Scientific Software - Peer-reviewed - Python
Published by tjkessler over 7 years ago

ECNet - Bug fixes, tool additions

Fixed bug where Server.train() did not obtain shuffled sets from its DataFrame if the user is shuffling sets for each candidate neural network
Server._sets is now public instead of private
Removed unnecessary import in ecnet\model.py
Added a suite of tools for creating an ECNet-formatted database:
- tools\createecnetdb.py (aggregates SMILES, QSPR descriptors for database creation)
- tools\nametosmiles.py (uses PubChemPy to obtain SMILES strings for supplied molecule names)
- tools\smilestoqspr.py (generates QSPR descriptors from SMILES strings using Open Babel and PaDEL-Descriptor)
- added version of PaDEL-Descriptor
- added how-to-use documentation (tools\README.md)

Scientific Software - Peer-reviewed - Python
Published by tjkessler over 7 years ago

ECNet - Multiprocessing and More

Rework of Server method arguments, functionality
- This includes an API change (check your scripts!)
Rework of Server training/selection algorithms
Improvements to logging
Rework of model saving/loading
When input dimensionality is reduced, selected inputs are applied to current data immediately
Updates to hyperparameter tuning
- Can now shuffle data sets for each bee
- Supports ECabc 2.1.0

Scientific Software - Peer-reviewed - Python
Published by tjkessler over 7 years ago

ECNet - Logging, DocStrings and Documentation

Added support for logging with the ColorLogging package
Updated docstrings, uses Google's documentation format
Modules now only import functions/classes they use instead of whole packages
Added unit testing for Server methods
Updated genetic limiting per PyGenetics 0.5.0 update
Made project variables accessible
Updated documentation
- Moved Server method documentation to ecnet/README.md
- Updated README.md to include more code block examples
- Updated terminology, project scope, clarity

Scientific Software - Peer-reviewed - Python
Published by tjkessler over 7 years ago