In the era of big data, can neural-network-based lossless compressors make storage more efficient?

As various types of data grow explosively, large-scale data storage, backup, and transmission become challenging, which motivates many researchers to propose efficient universal compression algorithms for multi-source data. In recent years, due to the emergence of hardware acceleration devices such as GPUs, TPUs, DPUs, and FPGAs, the performance bottleneck of neural networks (NN) has been overcome, making NN-based compression algorithms increasingly practical and popular. However, the research survey for the NN-based universal lossless compressors has not been conducted yet, and there is also a lack of unified evaluation metrics. To solve the problems, a research team led by Wang Gang, Liu Xiao-guang and Zhong Cheng published their new research on 15 July 2025 in Frontiers of Computer Science co-published by Higher Education Press and Springer Nature.

The team first divide NN-based algorithms into three types (static, adaptive, and semi-adaptive) and compare their advantages and weaknesses at a high level. Then, they introduce representative algorithms for each type, and provide pseudocode descriptions. Finally, they offer a summary of NN-based compressors.

In the research, they conduct experiments more than 4600 CPU/GPU hours to evaluate 17 state-of-the-art compressors on 28 real-world datasets (with a total data size of 8482 MB) across data types of text, images, videos, audio, etc. They also summarize the strengths and drawbacks of NN-based lossless data compressors and discuss promising research directions. They summarize the results as the NN-based Lossless Compressors Benchmark (NNLCB at https://fahaihi.github.io/NNLCB), which will be updated and maintained continuously in the future.