Estimating fault masking using squeeziness based on Rényi's entropy
Alfredo Ibias, Manuel A. Nunez · 2020
Squeeziness is an Information Theory notion that has been proven to strongly correlate with the likelihood of Failed Error Propagation (FEP). This allows us to estimate the FEP of a certain system by computing its Squeeziness. The original notion of Squeeziness is based on the classical notion of entropy defined by Shannon. In this paper we study alternative notions of Squeeziness based on a more general notion of entropy introduced by Rényi. In contrast to Shannon's entropy, which is univocally defined, Rényi's entropy depends on a parameter α. We define Squeeziness by using Rényi's entropy and analyse the correlation of the different notions of Squeeziness with the likelihood of FEP. Our experiments showed that although α = 1, corresponding to Shannon's entropy, induces good correlations, there are values of α showing better correlations.