Hierarchical Temporal Memories prediction performance and robustness to faults on multivariate time series

Mathieu Jégou (1, 2) , Pierre Chevaillier (3, 2) , Pierre de Loor (2)

1 ENIB - École Nationale d'Ingénieurs de Brest
2 Lab-STICC_ENIB_CID_IHSEV
3 LISYC - Laboratoire d'Informatique des Systèmes Complexes

In this article, we evaluate the ability of Hierarchical Temporal Memories (HTM) to process values coming from sensor chains. We present a study on the impact of the HTM parameterization on its ability to predict input values and its robustness to sensor faults. The HTM is evaluated on simulated multivariate time series comprising several causal relations between variables. The results show the ability of HTM to predict future values of multivariate time series and to be robust to sensor faults. We then present which parameters most impact HTM prediction performance and its robustness to faults.

Mots clés: Multivariate Time Series; Prediction; Fault Ro- bustness; Hierachical Temporal Memory

Domaines: Intelligence artificielle [cs.AI]

Mathieu Jégou, Pierre Chevaillier, Pierre de Loor. Hierarchical Temporal Memories prediction performance and robustness to faults on multivariate time series. 18th International Conference on Machine Learning and Applications (ICMLA 2019), special session on Machine Learning for Predictive Models in Engineering Applications, Dec 2019, Boca Raton, FL, United States. ⟨hal-02447618⟩