In this exercise, you will experience a few machine learning methods for design of new materials starting from an existing database. In particular, you will design a model to identify good candidate materials for light harvesting, based on a small database of organic/inorganic perovskites. Afterwards, you will validate the model predictions by running DFT calculations on selected systems.
Part 1: Inspection of database¶
The first part of the exercise is an inspection of the existing database. Understanding what is available from other sources is a necessary step before running any machine learning tools. Here, you will:
- Extract structures from the database.
- Calculate heat of formations.
- Plot histograms and scatter plots for different quantities available from the database.
Part 2: Machine Learning¶
In this part, you will implement the machine learning model:
- Define the input vectors.
- Select a suitable functional form with optimal parameters.
- Find a loss function to evaluate the performances.
- Apply this model to the prediction of the heat of formation.
- Improve the input vectors and the model.
Part 3: Test and Evaluate the Model¶
In the last part, you will test the prediction model and run DFT calculations for the heat of formation and the band gap to compare these results with the model.