Experiments on yield prediction

This repository contains the code to carry out experiments on yield prediction.

Dependencies

All dependencies are handled through Anaconda and were developed with Anaconda 4.10.1. After installing anaconda, initialize the environment for this repository using:

$> conda env create -f environment.yml

After this is done, initialize the environment by running:

$> conda activate yield-prediction

Code organization

The experiments are organized around python scripts. Experiments are configured using python script files (see the directory configs/). The script to run experiments is scripts/run_experiment.py. Run it from a command line like so:

$> python scripts/run_experiment.py configs/yield_prediction_experiment.py

All settings for experiments are configured in this file, which in turn will use other configuration files for the different models to run.

Explaining datasets

The code uses a modular approach to datasets. Datasets are described using a special python file containing an instance of the DatasetSpec class which describes to the experiment code how the dataset should be loaded. The experiment configuration file refers to this python file. To add a new dataset, you can copy one of the existing files (see datasets/uspto/raw_dataset_conf.py for an example).