SMBJClassifier is a universal classifier for Single-Molecule Break Junction (SMBJ) experiments. It takes raw current traces measured from SMBJ and automatically classifies them among several (user-specific) targeting single molecules.
More information regarding the overall approach, methods and results can be found in our publication:
SMBJClassifier includes:
- A series of preprocessing steps to remove the majority of noises within raw current traces.
- Conversion of current traces into sampled conductance histograms, which are input data for the classifier.
- Machine learning models for classifications, based on combining an ensemble learning method XGBoost and a convolutional neural network
- Four different input feature representations: 1D and 2D conductance probability distributions, with and without averaging over the experimental parameters
This installation method is intended for users who sets up a Python environment without pipenv.
pip install --upgrade "git+https://github.com/ethanwyr/SMBJClassifier.git"
All dependencies will be installed into your selected environment with the above command. Dependencies can be found in the requirements.txt file.
This installation method is intended for users who sets up a Python environment with pipenv. pipenv allows users to create and activate a virtual environment with all dependencies within the Python project. For more information and installation instructions for pipenv, see https://pipenv.pypa.io/en/latest/.
git clone the SMBJClassifier github repository.
From the base directory of your local SMBJClassifier git repo, create a Pipfile.lock file from Pipfile using:
pipenv install requests
Activate the virtual environment using:
pipenv shell
Deactivate the virtual environment using:
exit
To import SMBJClassifier into your Python script. From the base directory of your local SMBJClassifier git repo, use:
import sys
sys.path.append('../SMBJClassifier')
from SMBJClassifier import DPP, modelTo use SMBJClassifier, please refer to our demo with usage examples.