feature-selection-via-autoregreesive-generation

This is the implementation code of the paper "Neuro-Symbolic Embedding for Short and Effective Feature Selection via Autoregressive Generation"

implementation

Step 1: download the data:

follow the instruction in /data/readme.md

Step 2: collect the training data

python3 xxx/code/baseline/automatic_feature_selection_gen.py --name DATASETNAME --choice REDUNDANCY_CHOICE --unsupervised IS_UNSUPERVISED

Step 3: generate the optimal feature subset

python3 xxx/code/ours/train_controller.py ---method_name MODEL_CONFIGURATION --task_name DATASETNAME --gen_num GENERATION_SET_NUM --batch_size 64 --epochs 1000 --lr 0.0001