Most of the diff is moving files around; main changes are in evaluation/generate_queries.py. To use, first generate a workload using load_small_data.sh into a .txt file, then call the new script on it to generate queries for it. 100 queries are generated, labels are chosen uniformly at random, and metric values are chosen from a normal distribution around each metric's mean with variance mean/2. Metric ops are limited to the non-equality ones. See top lines for hyperparamaeters.
Most of the diff is moving files around; main changes are in
evaluation/generate_queries.py
. To use, first generate a workload usingload_small_data.sh
into a .txt file, then call the new script on it to generate queries for it. 100 queries are generated, labels are chosen uniformly at random, and metric values are chosen from a normal distribution around each metric's mean with variance mean/2. Metric ops are limited to the non-equality ones. See top lines for hyperparamaeters.