Medical Researchers around the world exploit the properties of the molecules to understand its structure and chemi- cal behavior which helps them to design molecules to carry out specific task and produce new drugs to fight deadly disease. The knowledge of magnetic interactions between the atoms of molecule i.e. Scalar Coupling Constant helps them to infer various properties of molecules. However, calculat- ing these interaction via tradition quantum mechanics methods is time consuming and computational extensive, which opens up the research for other predicting methodologies like statistical machine learning and deep learning algorithms. In addition, the domain of variables involves could be narrowed down via feature selection algorithms and transfer learning.
-
Data_Preprcessing.ipynb: This notebook consists of code for Data Exploration & Wrangling to create a final dataset required for training various Neural Network Architecture and Machine Learning models.
-
Neural_Net_Training.ipynb: Resulting
dataset.csvfile is being used in this notebook for various neural network architectures. Moreover, this notebook also comprises of multifarious experiments and respective visual results as discussed in project paper. -
Machine_Learning_Models.ipynb: This notebook hosts the code required for conducting machine learning experiements with the data.
- It is advisable to first dowload the data from here
Data_Preprocessing.ipynbshould be first executed which will produce the pre-processed data csv file 'dataset.csv'- Remaining two notebooks could be executed in any order. All the relevant results and graphs visualization are present in the respective jupyter notebook files.
Note: The entire startegy & methodlogies of the project are discussed in this video