Gama: General Adaptive Memory Allocation for Learned Bloom Filters

---

Gama is a holistic framework that solves the memory allocation problem of LBFs. The objective of Gama is to minimize the overall FPR of an LBF under a given memory budget

Prerequisites for testing

• Default Python version used in experiments: 3.10.12
• Run model_generate.py to generate a fixed-size LightGBM model for method comparison
• Run dataset/get_sample.py to generate the sampled dataset

Run

• We use URL Dataset as the default for testing

Method Name	Python Script Path
Gama-LBF	lgb_url_GamaLBF_ada.py
Gama-PLBF	lgb_url_GamaPLBF_ada.py
PLBF	PLBF/main.py
Ada-BF	ada-bf/main.py
Sandwich-BF	sandwich-lbf/main.py
LBF	lbf/main.py
BF	bf_url_main.py
LBF+GA	lgb_url_GamaLBF_ga.py
PLBF+GA	lgb_url_GamaPLBF_ga.py

Datasets

The datasets used in paper can be downloaded from the following links:

• URL：https://www.kaggle.com/datasets/sid321axn/malicious-urls-dataset
• Yelp: https://www.yelp.com/dataset/download
• COD: https://www.kaggle.com/datasets/hari31416/celestialclassify