A collection of benchmarks for PROBABILISTIC INFERENCE with algebraic and logical constraints.
The paper features experiments where Density Estimation Trees (DETs) are trained on UCI data (in data/mlc-datasets
). A number of queries over the continuous variable models are generated:
Go to uai-22/
and run:
python3 generate_dets.py N_MIN N_MAX NQUERIES QUERYHARDNESS SEED
where:
NMIN
andNMAX
are hyperparameters of the greedy DET learning algorithm. They constrain the min. and max. number of instances in the leaves of the DETNQUERIES
andQUERYHARDNESS
respectively control how many queries over the learned models are generated and the ratio of variables involved in the query.SEED
sets a seed number