Newest papers

Turning TS2Vec into a Foundation Model for Time Series Classification

Vasilii Feofanov, Zhenghan Tai, Andrei Ermakov, Irina Rish · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28602 · cited by 0 works (OpenAlex) · open-access PDF

Spanish

TS2Vec es uno de los métodos más fuertes de extracción de probabilidad para la clasificación de series de tiempo, introducido en 2022. TS2Vec está diseñado para ser entrenado de nuevo en cada conjunto de datos: pre-entrenamiento auto-supervisado seguido de probado lineal. Desde entonces, el campo se ha desplazado hacia modelos de fundación que reutilizan un codificador en todos los conjuntos de datos. Sin embargo, TS2Vec ofrece una arquitectura convolutiva compacta que funciona directamente en muestras temporales y acepta entradas de longitud variable sin reemplazar. Preguntamos si estas propiedades se pueden mantener cuando el codificador es pre-entrenado sólo una vez, y muestran que pueden: TS2Vec esencialmente se puede convertir en un modelo de fundación. Tres cambios son sufici

Machine translation (Meta M2M100); the English original follows.

Abstract

TS2Vec is one of the strongest self-supervised methods for time series classification, introduced in 2022. It is designed to be trained anew on every dataset: self-supervised pre-training followed by linear probing. Since then, the field has shifted towards foundation models that reuse one encoder across datasets. Yet TS2Vec offers a compact convolutional architecture that operates directly on temporal samples and accepts variable-length inputs without resampling. We ask whether these properties can be retained when the encoder is pre-trained only once, and show that they can: TS2Vec can essentially be turned into a foundation model. Three changes suffice: pre-train on a synthetic corpus, feed the encoder a raw amplitude view and an instance-normalized shape view of the input through two small stems, and distill one specialist encoder per view into a single encoder. The result, TS2Vec-FM, is a 1.6M-parameter dilated convolutional network that, kept frozen, reaches 83.3% on UCR-128 and 75.3% on UEA-27 under linear probing, within 0.3 pp of MantisV2, the current state of the art, and above TS2Vec trained separately on each target dataset. What we find most interesting is how different TS2Vec-FM is from other foundation models in design: near-parity is reached without attention, without a patch tokenizer, and without intermediate-layer extraction, using an "old-school" convolutional encoder and a two-layer MLP stem.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Encoder, Computer science, Foundation (evidence), Artificial intelligence, Field (mathematics), Reuse

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange