LISA is a learned index structure for spatial data, similar in functionality to a R-Tree. More details about LISA can be found in our SIGMOD 2020 paper.
We implement LISA with python 2.7 on the Ubuntu opearting system. You need to install two scientific computing package: numpy and scipy. The codes can be found here.
The input data for building LISA is a numpy ndarray file whose shape is NxD where D and N are the dimension and the number of the input keys, respectively.
We prepared three 2-dimensional datasets and three 3-dimensional datasets for validating our approach. 2d_uniform_data_0.npy / 3d_uniform_data_0.npy is used for building LISA; 2d_uniform_data_2.npy / 3d_uniform_data_2.npy and 2d_uniform_data_3.npy / 3d_uniform_data_3.npy are used for analyzing how LISA performs if we insert and delete keys, respectively.
The details of the hardwares we used for the experiments are shown as follows.
- Processor: 32 Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz
- L1d cache: 32K
- L1i cache: 32K
- L2 cache: 256K
- L3 cache: 20480K
- Memory: 128GB
- Secondary storage: HDD 3.7 TB
- Install the scientific computing packages:
- pip install numpy; pip install scipy; pip install scikit-learn;
-
Clone the repository under a directory
$Workspace$ . -
Download 2d_uniform_data_0.npy, 2d_uniform_data_2.npy, 2d_uniform_data_3.npy and 2d_uniform_query_ranges.qr. Put them in the directory
$Workspace$ /LISA/2d_uniform/data/. Download 3d_uniform_data_0.npy, 3d_uniform_data_2.npy, 3d_uniform_data_3.npy and 3d_uniform_query_ranges.qr. Put them in the directory$Workspace$ /LISA/3d_uniform/data/. -
cd
$Workspace$ /LISA/src and run main.py
The LISA is saved in the same directory with the corresponding dataset. If you want to test LISA on other datasets, you need to create a new directory
We will update the repo if any modifications are made in the future.