LAL

Code for paper Ksenia Konyushkova, Raphael Sznitman, Pascal Fua 'Learning Active Learning from Data', NIPS 2017

This code can be run with Jupyter notebook 'AL experiments'. You will need the following packages: numpy, sklearn, matplotlib, scipy, time, scipy, math, pickle. 'AL experiment' guides you through the main steps. It uses classes from folder ./Classes, data for the experiments is stored in ./data, data for learning a strategy is stored in ./lal datasets and the results are written into ./exp. Class ActiveLearner implements methods Random, Uncertainty Sampling and LAL. For more details, refer to the paper and comments in the code.