Description

Students write their own neural network code from scratch using Python and use it to predict chemical properties of molecules. The performance is compared to standard machine learning packages such as scikit-learn, Keras, and DeepChem.

 

Knowledge:

Basic principles behind Python programming, machine learning, and cheminformatics. Classification and regression using neural networks. Activation functions, back propagation using gradient descent, Overfitting, regularisation, hyperparameter optimisation, and training/validation/test sets.  SMILES strings, molecular fingerprints, and graph convolution as applied to molecules.

Skills:

Data manipulation and visualisation using Pandas, numpy, and Matplotlib/Seaborn. Manipulation of chemical data using RDKit. Use of scikit-learn, Keras, and DeepChem. 

Competences:

Prediction of chemical properties using machine learning. Critical evaluation of machine learning models.

Recommended qualifications

First year organic chemistry and mathematics