The paper presents TWET, a Time‑Warping Estimation Trainable model that estimates time‑warping functions from a single observation. It frames the problem as a stationarization task in the wavelet domain and uses a hierarchical dilated convolutional architecture with a differentiable stationarity criterion for end‑to‑end training. Experiments demonstrate that TWET achieves higher deformation reconstruction accuracy and faster computation than existing methods, enabling low‑latency applications.
By Corentin Presv\^ots (Phys-ENS), Adrien Meynard (Phys-ENS)
The paper introduces FastCDTW, an efficient approximation for Continuous Dynamic Time Warping (CDTW). It offers two computational strategies: an exact closed‑form solution for piecewise‑linear warping functions and an unbiased Monte Carlo estimator for smooth warping functions such as those implemented by implicit neural networks. Both approaches provide gradients with respect to the warping function and the input time series, making the algorithm GPU‑friendly and suitable for integration into complex learning tasks like time‑series averaging, classification, and deep domain adaptation.
By Pierre Houedry, Romain Tavenard, Nicolas Courty, Romaric Gaudel
arXiv:2606. 23472v2 Announce Type: replace-cross Abstract: Time series classification involves learning a mapping from a continuous, temporally ordered sequence of real-valued observations to discrete response variables, like class labels.
By Vicky Geneva Haney, Kamel Lahouel, Victor Rielly, Bruno M. Jedynak
The paper introduces TimeFT, a new tokenizer for time series that operates in the frequency domain to avoid spectral distortions caused by traditional subsequence partitioning. It uses frequency-domain partitioning, frequency shifting, and Nyquist sampling, is parameter‑free, and adds negligible computational overhead. Experiments across forecasting, classification, and anomaly detection show consistent and significant performance gains over existing tokenizers.
By Zelong Tian, Ming Jin, Bo Du, Shirui Pan
WinoTS introduces a wavelet‑based self‑distillation framework for time‑series models that uses time‑frequency augmentations to create multi‑scale structural views, avoiding distortion of signal dynamics. The method outperforms state‑of‑the‑art baselines in long‑term forecasting, cross‑domain zero‑shot transfer, and unsupervised anomaly detection, and linear probing on frozen representations often beats fully supervised training from scratch. Ablation studies show WinoTS is architecture‑agnostic and demonstrates that time‑frequency transformations offer a principled alternative to vision‑style spatial augmentations.
By Noam Major, Kathy Razmadze, Yoli Shavit
The paper introduces a spectrally-aligned latent-flow model for time‑series generation that trains the latent space to preserve dynamical properties relevant to synthetic data quality. By incorporating fine‑tuning losses based on Fourier, wavelet, and signature transforms, the method mitigates spectral mismatches caused by latent compression and ensures alignment with true signals in terms of smoothness and targeted spectral content. Experiments on real‑world long‑range univariate and multivariate benchmarks show that the aligned model outperforms a base latent‑flow model and state‑of‑the‑art approaches in signal realism, computational efficiency, and local structure alignment.
By Camilo Carvajal Reyes, Felipe Tobar
WaveletDiff is a diffusion-based framework that generates time series by training directly on wavelet coefficients, leveraging multi-resolution structure through level‑specific transformers and cross‑level attention with adaptive gating. The model incorporates Parseval‑theorem‑based energy constraints to preserve time‑frequency properties during diffusion. Experiments on six real‑world datasets from energy, finance, and neuroscience show that WaveletDiff outperforms several diffusion baselines and competes with the VAE/transformer‑based MSDformer, achieving lower discriminative and Context‑FID scores while using fewer parameters and less training time.
By Yu-Hsiang Wang, Olgica Milenkovic
The paper introduces a new framework for learning time‑varying graphs from spatiotemporal data, leveraging a prior on signal temporal behavior to estimate graphs from few measurements. It adds three convex regularization terms that enforce sparsity in the temporal changes of the network, and presents a scalable algorithm to solve the resulting optimization problem. Experiments on synthetic data and real datasets—including point clouds, temperature readings, and EEG signals—show that the method outperforms existing state‑of‑the‑art approaches.
By Haruki Yokota, Koki Yamada, Yuichi Tanaka, Antonio Ortega
The paper introduces XACT, a framework that learns sparse attribution masks over coefficients from any invertible time‑frequency transform, such as STFT, continuous wavelet transform, and discrete wavelet transform. XACT extends the virtual inspection layer approach to wavelet transforms, enabling Layer‑wise Relevance Propagation (LRP) to generate explanations in these representations. Experiments on synthetic and two real‑world datasets show that XACT produces precise, sparse, and structured explanations, often outperforming baseline methods in highlighting relevant features.
By Theresa Dahl Frehr, Francisco Pelayo, Lukas Raad, Alicia Garc\'ia Sanz, Thea Br\"usch, Tommy Sonne Alstr{\o}m
arXiv:2608.24027v1 Announce Type: new
Abstract: 4D medical image interpolation aims to recover missing volumes from sparsely observed time points and is important for dynamic anatomical analysis in a...
By Haojin Li, Hengzhuo Wang, Zhiheng Ma, Mingyang Ou, Heng Li, Jiang Liu
The paper introduces cylindrical geodesic flow matching, a method for translating quasiperiodic cardiovascular waveforms between different body locations. By replacing the standard affine path in flow matching with a closed‑form geodesic on a phase–amplitude cylinder, the approach preserves amplitude and instantaneous frequency during interpolation. Experiments on photoplethysmography and seismocardiography data show that this geometry‑aware method outperforms interpolation baselines and matches or exceeds supervised models, reducing error metrics by up to ~15%.
By Onur Selim Kilic, Afra Nawar, Cem Okan Yaldiz, Michael J. Cho, Ahmet Rasim Emirdagi, Demet Tangolar, Amirali Aghazadeh, Amit J. Shah, Omer T. Inan
arXiv:2609.34965v2 Announce Type: replace-cross
Abstract: Many physiological time series, such as cardiac and brain recordings, exhibit cyclostationarity: their statistics vary periodically with an u...
By Samuel Ruiperez-Campillo, Michele Copetti, Jorge da Silva Goncalves, Sonia Laguna, Thomas Hofmann, Julia E. Vogt