# Exact, Parallelizable Dynamic Time Warping Alignment with Linear Memory

arXiv: https://arxiv.org/abs/2008.02734 · HTML: https://mlsanity.com/paper/2008.02734/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 4 Aug 2020.

## Built on by (most cited first, 3 of 3)

- [Online Symbolic Music Alignment with Offline Reinforcement Learning](https://mlsanity.com/paper/2401.00466/) (2023)
- [Latent Optimal Paths by Gumbel Propagation for Variational Bayesian Dynamic Programming](https://mlsanity.com/paper/2306.02568/) (2023)
- [Intelligently Augmented Contrastive Tensor Factorization: Empowering Multi-dimensional Time Series Classification in Low-Data Environments](https://mlsanity.com/paper/2505.03825/) (2025)

## Abstract (arXiv)

Audio alignment is a fundamental preprocessing step in many MIR pipelines. For two audio clips with M and N frames, respectively, the most popular approach, dynamic time warping (DTW), has O(MN) requirements in both memory and computation, which is prohibitive for frame-level alignments at reasonable rates. To address this, a variety of memory efficient algorithms exist to approximate the optimal alignment under the DTW cost. To our knowledge, however, no exact algorithms exist that are guaranteed to break the quadratic memory barrier. In this work, we present a divide and conquer algorithm that computes the exact globally optimal DTW alignment using O(M+N) memory. Its runtime is still O(MN), trading off memory for a 2x increase in computation. However, the algorithm can be parallelized up to a factor of min(M, N) with the same memory constraints, so it can still run more efficiently than the textbook version with an adequate GPU. We use our algorithm to compute exact alignments on a collection of orchestral music, which we use as ground truth to benchmark the alignment accuracy of several popular approximate alignment schemes at scales that were not previously possible.