Soft-DTW: a Differentiable Loss Function for Time-Series
arXiv:1703.01541
Abstract
We propose in this paper a differentiable learning loss between time series, building upon the celebrated dynamic time warping (DTW) discrepancy. Unlike the Euclidean distance, DTW can compare time series of variable size and is robust to shifts or dilatations across the time dimension. To compute DTW, one typically solves a minimal-cost alignment problem between two time series using dynamic programming. Our work takes advantage of a smoothed formulation of DTW, called soft-DTW, that computes the soft-minimum of all alignment costs. We show in this paper that soft-DTW is a differentiable loss function, and that both its value and gradient can be computed with quadratic time/space complexity (DTW has quadratic time but linear space complexity). We show that this regularization is particularly well suited to average and cluster time series under the DTW geometry, a task for which our proposal significantly outperforms existing baselines. Next, we propose to tune the parameters of a machine that outputs time series by minimizing its fit with ground-truth labels in a soft-DTW sense.
arXiv commentPublished in ICML 2017
References in corpus (5)
Cited by in corpus (17)
- Boosting Continuous Sign Language Recognition via Cross Modality Augmentation
- GDPNet: Refining Latent Multi-View Graph for Relation Extraction
- ECGadv: Generating Adversarial Electrocardiogram to Misguide Arrhythmia Classification System
- Fused Gromov-Wasserstein distance for structured objects: theoretical foundations and mathematical properties
- Dynamic Time Warping Clustering to Discover Socio-Economic Characteristics in Smart Water Meter Data
- Parallel Tacotron 2: A Non-Autoregressive Neural TTS Model with Differentiable Duration Modeling
- Causal Curiosity: RL Agents Discovering Self-supervised Experiments for Causal Representation Learning
- Distribution Regression for Sequential Data
- Learning Registered Point Processes from Idiosyncratic Observations
- Multi-rate attention architecture for fast streamable Text-to-speech spectrum modeling
- Towards Diverse Perspective Learning with Selection over Multiple Temporal Poolings
- Learning Discriminative Prototypes with Dynamic Time Warping
- Vector quantisation and partitioning of COVID-19 temporal dynamics in the United States
- Trainable Time Warping: Aligning Time-Series in the Continuous-Time Domain
- Video2Skill: Adapting Events in Demonstration Videos to Skills in an Environment using Cyclic MDP Homomorphisms
- End-To-End Dilated Variational Autoencoder with Bottleneck Discriminative Loss for Sound Morphing -- A Preliminary Study
- Online Graph Dictionary Learning