Time Series Source Separation with Slow Flows

Edouard Pineau (1, 2, 3) , Sébastien Razakarivony (1) , Thomas Bonald (2, 4, 5)

1 Safran Tech
2 IP Paris - Institut Polytechnique de Paris
3 Télécom Paris
4 INFRES - Département Informatique et Réseaux
5 DIG - Data, Intelligence and Graphs

In this paper, we show that slow feature analysis (SFA), a common time series decomposition method, naturally fits into the flow-based models (FBM) framework, a type of invertible neural latent variable models. Building upon recent advances on blind source separation, we show that such a fit makes the time series decomposition identifiable.

Domaines: Apprentissage [cs.LG]

Edouard Pineau, Sébastien Razakarivony, Thomas Bonald. Time Series Source Separation with Slow Flows. ICML Workshop on Invertible Neural Networks, Normalizing Flows, and Explicit Likelihood Models, 2020, (On line ), France. ⟨hal-02959282⟩