/CAFEM

Source code for "Cross-data Automatic Feature Engineering via Meta-learning and Reinforcement Learning"

Primary LanguagePython

Source code for "Cross-data Automatic Feature Engineering viaMeta-learning and Reinforcement Learning"

Introduction

Packages

CAFEM relays on several common packages:

tensorflow

sklearn

numpy

multiprocessing

tqdm

Training

Train FeL with Order-1 Operators on 20 datasets

for val in 620 589 586 618 616 607 795 770 1480 913 735 1067 917 4154 1049 806 1566 954 1038 825
do
    python single_afem.py --num_epochs 100 --dataset $val --cuda 0 --n_jobs 1 --buffer_size 1000 --num_episodes 1 --out_dir ../out/skip_$val 
done

Train FeL with Order-1 & Order-2 operators on 20 datasets

Train CAFEM on 100 datasets and apply on each of other 20 datasets

python cafem.py --multiprocessing 6 --out_dir ../out/ml --num_episodes 5

python single_afem.py --load_weight ../out/ml/model/model_5.ckpt --dataset 1049 --out_dir ../out/o2_5_1049 --num_epochs 50 --buffer_size 1000 --num_episodes 1