MomentQuant: an even more minimalist interval method with linear time complexity for time series classification

On September 4 and 7, 2026, arXiv published papers related to MomentQuant. The study introduced a minimal interval method called MomentQuant, designed for time series classification tasks. The core improvement of this method lies in achieving linear time complexity, primarily by optimizing the precise implementation of the Quant algorithm and using Cornish-Fisher expansion to derive approximate quantiles, thereby eliminating the need for sorting. Experimental results show that MomentQuant outperforms the existing Quant implementations and its own version in推理 speed, although its prediction performance has a slight decline.