[LAMPS] A signature-based functional spatial autoregressive model

Résumé

We propose a new approach to the spatial autoregressive model with functional covariates, based on the concept of signature. This represents a function as an infinite series of its iterated integrals and has the advantage of being applicable to a wide range of processes. After providing theoretical guarantees for the proposed model, we showed, in a simulation study and on a real dataset, that this new approach offers competitive performance compared to the conventional model.
 

Intervenant