RTree Learned Index

This is the main code setup used to carry out Fall 2019 Research on Learned Indexes for R-Trees.

Dependencies

• Sklearn
• Pytorch
• Numpy
• Pandas
• pyrtree (needs to be installed)

Running Scripts

All the training scripts are present within pyrtree/tests

Building Dataset

Run buildset to build different types of datasets. Choices are present in argparse. Either path or LeafID can be produced. Rectangle distribution can be modified in test_rtree.

Training Models

Hybrid Model

The hybrid Rtree model can be trained in train_hybrid. Tree parameters and approximators can be modified within the script.

Neural Network

Direct Mapping using Neural Network can be carried out using the NN/ directory. This requires pytorch. Make sure to normalize the dataset before running.

Rtree Implementation Source

https://github.com/Rhoana/pyrtree Double check the implementation (I had a couple issues pop up that I wasn't able to pursue.)