Identifying Easy Instances to Improve Efficiency of ML Pipelines for Algorithm-Selection - Code and Data

This repository contains the code and data for reproducibility of the paper 'Identifying Easy Instances to Improve Efficiency of ML Pipelines for Algorithm-Selection'. 

The following files are included:

best_algo.csv : labels for the classification and median performance of algorithms;

ML_models.ipyb : jupyter notebook with the definition of the neural networks for both classifiers;

pickle.zip : pickled models for the hardness classification and the algorithm selector;

trajectories.zip : raw data files containing parts of the trajectories of each algorithm;

Results.zip : results obtained using the approach on the stream of instances.