A Deep Dynamic Latent Block Model for the Co-Clustering of Zero-Inflated Data Matrices

Abstract

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. Copyright

Deep dynamic co-clustering of streams of count data: a new online Zip-dLBM

“…The following section reminds the Zero-Inflated Poisson Dynamic Latent Block Model (Zip-dLBM, Marchello et al, 2022a) for batch processing. This paper has been proposed in a general way for any Zero-Inflated distribution, however, here we focus on the Zero-Inflated Poisson (ZIP) distribution.…”

“…The terms on the right hand side of the above equation can be further developed, for details see Marchello et al (2022a).…”

“…We denote by δ ij (t) := q(A ij (t) = 1) the variational probability of success for A ij (t), τ iq (t) := q(Z iq (t) = 1) the variational probability of success of Z iq (t), and η j (t) := q(W j (t) = 1) the variational probability of success of W j (t). The explicit updating equations and the proofs are provided in Marchello et al (2022a).…”

“…Furthermore, there is a growing need to develop machine learning models for time-dependent sparse data matrices and, although many notable methods have been introduced in this field in recent decades (Marchello et al, 2022b;Casa et al, 2021), the development of online co-clustering methods remains largely unexplored. This paper proposes an online extension of Zip-dLBM (Marchello et al, 2022a). We introduce three novelties in this regard.…”