NeaTS is a learned compressor for time series providing, simultaneously, compression ratios close to or better than the best existing compressors, a faster decompression speed, and orders of magnitude more efficient random access.

• 🧩 Piecewise nonlinear approximations
• 🚀 Efficient random access and range queries without full decompression
• 🧠 SIMD-accelerated decompression (AVX2/AVX-512)
• 📉 Lossless and lossy modes

• C++23 compatible compiler (e.g., GCC ≥ 13 or Clang ≥ 16)
• CMake ≥ 3.22
• SDSL-lite
• Squash 0.7+ (required for benchmarking)

git clone https://github.com/and-gue/NeaTS.git
cd NeaTS
mkdir build && cd build
cmake ..
make -j$(nproc)

./DecompressorSIMD path/to/data.bin <bpc>

• data.bin — Binary file of 64-bit signed integers
• <bpc> — Maximum residual (size in bits)

The datasets used in the paper are available at this link: NeaTS Datasets

If you use NeaTS for research, please cite:

@inproceedings{guerra2025neats,
  author    = {Guerra, Andrea and Vinciguerra, Giorgio and Boffa, Antonio and Ferragina, Paolo},
  title     = {Learned Compression of Nonlinear Time Series with Random Access},
  booktitle = {2025 IEEE 41st International Conference on Data Engineering (ICDE)},
  year      = {2025},
  pages     = {1579--1592},
  doi       = {10.1109/ICDE65448.2025.00122}
}

This project is released for academic purposes under the terms of the GNU General Public License v3.0.