Published September 28, 2025 | Version 1.0.0
Publication Open

BIE: Bit-Index Encoding for Efficient Neural Network Weight Compression

Description

A novel compression framework that represents neural network weights through bit-level indexing, achieving compression ratios up to 40× with minimal accuracy degradation. BIE operates by encoding sparse weight matrices as sets of bit positions where non-zero values occur, providing three encoding variants: binary encoding for maximum compression, bitplane encoding for balanced compression-accuracy trade-offs, and blocked encoding for improved cache locality. The framework includes optimized sparse matrix multiplication kernels using Numba JIT compilation, comprehensive benchmarking tools, and integration capabilities with popular deep learning frameworks. All experiments are fully reproducible using the provided source code and datasets.

Files

BIE_Research_Paper.pdf

Files (2.4 MB)

Name Size Download all
md5:8666002c3fb4398c1af3c24e2d68d08d
2.4 MB Preview Download

Additional details

Related works

Is supplemented by
Software: https://github.com/GruheshKurra/bit-index-encoding-research (URL)

Software

Repository URL
https://github.com/GruheshKurra/bit-index-encoding-research
Programming language
Python
Development Status
Active

References

  • Jacob, B., Kligys, S., Chen, B., Zhu, M., Tang, M., Howard, A., Adam, H., & Kalenichenko, D. (2018). Quantization and training of neural networks for efficient integer-arithmetic-only inference. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2704-2713.
  • Wu, X., Yao, Z., & He, Y. (2023). ZeroQuant-FP: A Leap Forward in LLMs Post-Training W4A8 Quantization Using Floating-Point Formats. arXiv preprint arXiv:2307.09782.
  • Han, S., Mao, H., & Dally, W. J. (2015). Deep compression: Compressing deep neural networks with pruning, trained quantization and huffman coding. arXiv preprint arXiv:1510.00149.
  • Frankle, J., & Carbin, M. (2019). The lottery ticket hypothesis: Finding sparse, trainable neural networks. International Conference on Learning Representations.
  • Chen, Y., Li, J., Xiao, H., Jin, X., Yan, S., & Feng, J. (2022). A comprehensive survey of neural network compression. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(10), 5436-5454.
  • Nagel, M., Fournarakis, M., Amjad, R. A., Bondarenko, Y., van Baalen, M., & Blankevoort, T. (2021). A white paper on neural network quantization. arXiv preprint arXiv:2106.08295.
  • Gale, T., Elsen, E., & Hooker, S. (2019). The state of sparsity in deep neural networks. arXiv preprint arXiv:1902.09574.
  • Louizos, C., Welling, M., & Kingma, D. P. (2018). Learning sparse neural networks through L0 regularization. International Conference on Learning Representations.
  • Dettmers, T., Svirschevski, R., Egiazarian, V., Kuznedelev, D., Frantar, E., Ashkboos, S., Borzunov, A., Hoefler, T., & Alistarh, D. (2024). SpQR: A sparse-quantized representation for near-lossless LLM weight compression. International Conference on Learning Representations.
  • Lin, J., Tang, J., Tang, H., Yang, S., Dang, X., Gan, C., & Han, S. (2024). AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration. arXiv preprint arXiv:2306.00978.
  • Frantar, E., & Alistarh, D. (2023). SparseGPT: Massive language models can be accurately pruned in one-shot. International Conference on Machine Learning, 10323-10337.
  • Zhang, Y., Zhao, L., Cao, S., Zhang, S., Wang, W., Cao, T., Yang, Z., & Li, J. (2024). Integer or floating point? New outlooks for low-bit quantization on large language models. 2024 IEEE International Conference on Multimedia and Expo (ICME), 1-6.
  • Xu, Y., Han, X., Yang, Z., Wang, S., Zhu, Q., Liu, Z., Sun, M., & Li, P. (2024). OneBit: Towards extremely low-bit large language models. arXiv preprint arXiv:2402.11295.
  • Wang, H., Ma, S., Dong, L., Huang, S., Wang, H., Ma, L., Yang, F., Wang, R., Wu, Y., & Wei, F. (2023). BitNet: Scaling 1-bit transformers for large language models. arXiv preprint arXiv:2310.11453.
  • Huang, W., Liu, Y., Qin, H., Li, Y., Zhang, S., Liu, X., Magno, M., & Qi, X. (2024). BiLLM: Pushing the limit of post-training quantization for LLMs. arXiv preprint arXiv:2402.04291.